CN111080542B - Image processing method, device, electronic equipment and storage medium - Google Patents

Abstract

The present application proposes an image processing method, device, electronic device and storage medium, wherein the method comprises: obtaining a captured image, identifying a subject in the captured image, determining the degree of distortion of the subject if the subject is identified, and determining whether to perform de-distortion processing on the portrait area in the captured image according to the degree of distortion of the subject. Thus, by performing de-distortion processing only on the subject with a greater degree of distortion in the captured image, the original state of the captured image is retained to the greatest extent, the amount of image processing calculation is reduced, and the efficiency of de-distortion processing of the captured image is improved.

Description

Image processing method, device, electronic device and storage medium

Technical Field

The present application relates to the field of image processing technology, and in particular to an image processing method, device, electronic device and storage medium.

Background technique

At present, with the advancement of smart terminal manufacturing technology, smart terminals are equipped with camera modules for users to take photos, among which wide-angle cameras are more commonly installed on smart terminals. Compared with traditional lens cameras, wide-angle lens cameras have a larger field of view (FOV), but wide-angle lenses have larger distortion, and the edges of images will be seriously distorted.

In the related art, in order to compensate for the distortion of the image captured by the wide-angle camera, it is necessary to perform distortion correction processing on the image. Currently, the distortion correction processing on the image is performed on the entire captured image, which has the problem of low processing efficiency.

Summary of the invention

The present application aims to solve one of the technical problems in the related art at least to some extent.

The first embodiment of the present application provides an image processing method, including:

Acquire captured images;

Identifying a subject person in the captured image;

If the subject person is identified, determining the degree of distortion of the subject person;

According to the degree of distortion of the subject person, it is determined whether to perform dedistortion processing on the portrait area in the captured image.

As a first possible scenario of an embodiment of the present application, after identifying the subject person in the captured image, the method further includes:

If the subject person is not identified, querying the number of the portrait areas in the captured image;

If the number of the portrait areas is greater than or equal to a first threshold, the proportion of front faces in the faces presented in the portrait areas is counted;

If the proportion of the front face is greater than a second threshold, determining to perform a dedistortion process on the portrait area in the captured image;

If the proportion of the front face is less than or equal to or greater than the second threshold, it is determined that there is no need to perform dedistortion processing on the portrait area in the captured image.

As a second possible situation of the embodiment of the present application, after querying the number of the portrait areas in the captured image, the method further includes:

If the number of the portrait areas is less than the first threshold, it is determined that there is no need to perform dedistortion processing on the portrait areas in the captured image.

As a third possible situation of the embodiment of the present application, the identifying the subject person in the captured image includes:

Performing face recognition on the captured image;

Determine the face size, face rotation angle and face clarity for each face;

The face whose face size, face rotation angle and face clarity all meet the set conditions is taken as the main person in the captured image.

As a fourth possible situation of the embodiment of the present application, determining the degree of distortion of the subject person includes:

Determining the FOV of the subject person according to the position of the subject person in the captured image;

If it is determined that the FOV is less than the predetermined angle threshold, it is determined that the subject person has no distortion;

If it is determined that the FOV is greater than or equal to the angle threshold, predicting the real contour according to the imaging contour of the subject person;

The degree of distortion of the subject person is determined according to the difference between the imaging contour and the real contour.

As a fifth possible situation of the embodiment of the present application, after determining to perform a dedistortion process on the portrait area in the captured image, the method further includes:

Identifying straight line segments in the captured image;

The portrait region in the captured image is dedistorted according to the straight line segments in the captured image, so as to keep the straight line segments in the same shape before and after the dedistortion.

As a sixth possible situation of the embodiment of the present application, the identifying straight line segments in the captured image includes:

Determining a plurality of edge points from each pixel according to a gradient value of each pixel in the captured image and pixel values of adjacent pixels;

Fitting the multiple edge points to obtain multiple initial straight line segments; wherein each initial straight line segment is obtained by fitting edge points with similar gradient directions;

The multiple initial straight line segments are merged to obtain straight line segments in the captured image.

As a seventh possible situation of the embodiment of the present application, the fitting of the multiple edge points to obtain multiple initial straight line segments includes:

Determine multiple sets according to edge points with similar gradient directions among the multiple edge points; wherein the edge points in the same set have similar gradient directions;

For each set, the edge points in the corresponding set are fitted to obtain an initial straight line segment.

As an eighth possible situation of the embodiment of the present application, determining multiple sets according to edge points with similar gradient directions among the multiple edge points includes:

Determine the initial reference point from the edge points that have not been added to either set;

Querying edge points whose gradient direction difference with the reference point is less than an angle threshold and which are adjacent to the reference point;

Adding the queried edge points and the reference points to the same set;

If the degree of discreteness of the gradient direction of each edge point in the same set is less than or equal to the set discreteness, the queried edge point is used as the updated reference point to repeatedly execute the step of querying the edge point whose gradient direction difference with the reference point is less than the angle threshold and is adjacent to the reference point, and adding the queried edge point and the reference point to the corresponding set until the degree of discreteness of the gradient direction of each edge point in the corresponding set is greater than the set discreteness.

The image processing method of the embodiment of the present application obtains a captured image, identifies the subject person in the captured image, and if the subject person is identified, determines the degree of distortion of the subject person, and determines whether to perform de-distortion processing on the portrait area in the captured image according to the degree of distortion of the subject person. Thus, by only performing de-distortion processing on the subject person with a large degree of distortion in the captured image, the original state of the captured image is retained to the greatest extent, the amount of image processing calculation is reduced, and the efficiency of de-distortion processing of the captured image is improved.

The second aspect of the present application provides an image processing device, including:

An acquisition module, used for acquiring captured images;

A recognition module, used to recognize the subject person in the captured image;

A determination module, configured to determine the degree of distortion of the subject person if the subject person is identified;

The processing module is used to determine whether to perform de-distortion processing on the portrait area in the captured image according to the distortion degree of the subject.

The image processing device of the embodiment of the present application obtains a captured image, identifies the subject person in the captured image, and if the subject person is identified, determines the degree of distortion of the subject person, and determines whether to perform de-distortion processing on the portrait area in the captured image according to the degree of distortion of the subject person. Thus, by performing de-distortion processing only on the subject person with a large degree of distortion in the captured image, the original state of the captured image is retained to the greatest extent, the amount of image processing calculation is reduced, and the efficiency of de-distortion processing of the captured image is improved.

The third aspect of the present application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the image processing method as described in the above embodiment is implemented.

The fourth aspect of the present application provides a non-temporary computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the image processing method described in the above embodiment is implemented.

Additional aspects and advantages of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through the practice of the present application.

BRIEF DESCRIPTION OF THE DRAWINGS

The above and/or additional aspects and advantages of the present application will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

FIG1 is a schematic diagram of a flow chart of a first image processing method provided in an embodiment of the present application;

FIG2 is a schematic diagram of a flow chart of a second image processing method provided in an embodiment of the present application;

FIG3 is a schematic diagram of a flow chart of a third image processing method provided in an embodiment of the present application;

FIG4 is a schematic diagram of a fourth image processing method according to an embodiment of the present application;

FIG5 is a schematic diagram of a fifth image processing method according to an embodiment of the present application;

FIG6 is an exemplary diagram of an image processing method provided by an embodiment of the present application;

FIG7 is a schematic diagram of a sixth image processing method provided in an embodiment of the present application;

FIG8 is a schematic diagram of the structure of an image processing device provided in an embodiment of the present application;

FIG. 9 is a schematic structural diagram of an electronic device according to an embodiment of the present application.

Detailed ways

The embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present application, and should not be construed as limiting the present application.

In the related art, when performing de-distortion processing on a captured image, the entire captured image is usually corrected, which leads to a technical problem that the amount of computation required for the entire de-distortion process is relatively large.

In view of the technical problems in the related art, the present application proposes an image processing method, which obtains a captured image, identifies the subject in the captured image, and if the subject is identified, determines the degree of distortion of the subject, and determines whether to perform de-distortion processing on the portrait area in the captured image according to the degree of distortion of the subject. Thus, by only performing de-distortion processing on the distorted subject in the captured image, the original state of the captured image is retained to the greatest extent, the amount of image processing calculation is reduced, and the efficiency of de-distortion processing of the captured image is improved.

The following describes the image processing method, device, electronic device and storage medium of the embodiments of the present application with reference to the accompanying drawings.

FIG1 is a schematic flow chart of a first image processing method provided in an embodiment of the present application.

The embodiment of the present application takes the image-based image processing method being configured in an image processing device as an example. The image processing device can be applied to any electronic device so that the electronic device can perform an image processing function.

Among them, the electronic device may be a personal computer (PC), a cloud device, a mobile device, etc. The mobile device may be, for example, a mobile phone, a tablet computer, a personal digital assistant, a wearable device, a car device, and other hardware devices with various operating systems.

As shown in FIG1 , the image processing method comprises the following steps:

Step 101, obtaining a captured image.

In the embodiment of the present application, the captured image can be acquired by an image sensor provided in the electronic device.

As a possible scenario, the electronic device may include a visible light image sensor, and the image may be captured based on the visible light image sensor in the electronic device. Specifically, the visible light image sensor may include a visible light camera, and the visible light camera may capture visible light reflected by an imaging object for imaging.

As another possible situation, in an embodiment of the present application, the electronic device may further include a structured light image sensor, and images may be captured based on the structured light image sensor in the electronic device. Optionally, the structured light image sensor may include a laser lamp and a laser camera. Pulse Width Modulation (PWM) may modulate the laser lamp to emit structured light, and the structured light is irradiated to the imaging object. The laser camera may capture the structured light reflected by the imaging object for imaging, and obtain a structured light image corresponding to the imaging object.

It should be noted that the image sensor provided in the electronic device is not limited to the above-mentioned visible light sensor and structured light sensor, but may also be other types of image sensors, such as depth sensors, etc., which is not limited in the present application.

Step 102: Identify the subject person in the captured image.

In the embodiment of the present application, after the captured image is acquired, the subject person in the captured image is further identified.

As a possible implementation manner, the captured image may be input into a trained subject person recognition model to recognize the subject person in the captured image.

As another possible implementation, face recognition can be performed on the captured image first, and then the face size, face rotation angle and face clarity of each recognized face can be determined. Then, the face whose face size, face rotation angle and face clarity meet the set conditions can be regarded as the main person in the captured image.

Specifically, a face recognition model based on a convolutional neural network (CNN) can be used to perform face recognition on the captured image to identify each face in the captured image and determine the face size and position of each face. The face recognition model is trained using a large number of training sample images.

In the embodiment of the present application, face recognition is performed on the captured image, and after the face size and face position of each face are identified, face key point detection is performed on each face to determine the key points of the face, and then determine the key area position of the face. Among them, face key point detection is also called face key point detection, positioning or face alignment, which means locating the key area position of the face, including eyebrows, eyes, nose, mouth, facial contour, etc., given a face image. For example, 106 points of face key point detection can be performed on each face to obtain the facial features and contours of the face.

Furthermore, after determining the key points of each face, the face rotation angle of each face can be calculated based on the key points of each face. The face rotation angle refers to the face orientation information, such as whether the face is facing forward or sideways.

As a possible implementation method, the rotation parameters (roll, pitch, yaw) of the face can be calculated according to the key points of each face, so as to determine the rotation angle of each face.

In the embodiment of the present application, when determining the degree of facial clarity for each face, the face identified in the captured image can be first cropped to obtain each face image. Then, each face image is subjected to Gaussian blur denoising and converted into a grayscale image. The Laplace operator is used for filtering on each grayscale image, and the number of grayscale levels in each grayscale image is counted to obtain a grayscale histogram of each grayscale image. Each grayscale histogram is normalized and mapped to a range of 0 to 255, and the mean of the pixel values of each grayscale image after mapping is calculated. If the mean is greater than the set threshold, it is determined that the face in the face image is clear; if the mean is less than or equal to the set threshold, it is determined that the face in the face image is blurred.

In the embodiment of the present application, after determining the face size, face rotation angle and face clarity of each face, the face whose face size, face rotation angle and face clarity meet the set conditions is taken as the main person in the captured image.

For example, if the face size is smaller than the set size, it may be that the face is far away from the camera, and the face can be considered as the image background. For another example, if the determined face rotation angle is greater than the set angle threshold, it may be that the face is in profile, and the face can also be considered as the image background. For another example, if the face is blurry and the face clarity is less than the set clarity threshold, the face can be considered as the image background.

Step 103: If the subject person is identified, the degree of distortion of the subject person is determined.

It is understandable that when images are captured by an image sensor installed in an electronic device, due to image sensor performance errors, such as camera focal length changes, lens optical distortion, perspective errors during imaging, etc., the captured images are inevitably distorted.

In the embodiment of the present application, after the subject person is identified from the captured image, it is necessary to further determine the degree of distortion of the subject person, so as to determine whether to further perform de-distortion processing on the subject person according to the degree of distortion of the subject person.

As a possible implementation method, the offset angle of each subject character relative to the lens optical axis can be determined according to the position of each subject character in the captured image, and then whether the subject character is distorted can be determined according to the offset angle of each subject character. If the offset angle is less than an angle threshold, such as 60°, it is determined that the subject character is not distorted. If the offset angle is greater than or equal to the angle threshold, the real contour is further predicted based on the imaging contour of the subject character to compare the degree of difference between the imaging contour and the real contour. If the difference between the imaging contour and the real contour is large, it can be determined that the degree of distortion of the subject character is more serious; if the difference between the imaging contour and the real contour is small, or there is basically no difference, it can be determined that the subject character is not distorted.

Step 104: Determine whether to perform de-distortion processing on the portrait area in the captured image according to the degree of distortion of the subject person.

In an embodiment of the present application, after determining the degree of distortion of each subject person in the captured image, it can be determined whether de-distortion processing needs to be performed on the portrait area in the captured image based on the degree of distortion of the subject person.

As a possible implementation method, the determined distortion degree of each subject person can be compared with the distortion degree threshold according to the set distortion degree threshold. If the distortion degree of the subject person is greater than the distortion degree threshold, the portrait area corresponding to the subject person in the captured image is subjected to dedistortion processing; if the distortion degree of the subject person is less than the distortion degree threshold, the portrait area corresponding to the subject person in the captured image does not need to be subjected to dedistortion processing;

In an embodiment of the present application, when de-distorting a portrait area in a captured image, the portrait area can be first divided into a face area and a body area, and then the portrait area correction calculation is performed according to a preset initial projection grid to obtain a first correction size value corresponding to the face area and a second correction size value corresponding to the body area. Furthermore, among the first correction size value and the second correction size value, a target correction size value that meets preset conditions is determined, so that the portrait area in the image is corrected according to the target correction size value to obtain a de-distorted captured image.

When obtaining a first corrected size value corresponding to the face area and a second corrected size value corresponding to the body area, as a possible implementation method, an original grid in the portrait area can be constructed according to the coordinates of the pixel points in the portrait area, a preset initial projection grid is used to perform correction calculation on the portrait area, a first transformed grid corresponding to the face area and a second transformed grid corresponding to the body area are obtained, a size ratio of the first transformed grid to the original grid is calculated to obtain a first corrected size value, and a size ratio of the second transformed grid to the original grid is calculated to obtain a second corrected size value.

When obtaining the first corrected size value corresponding to the face area and the second corrected size value corresponding to the body area, as another possible implementation method, the depth value of the coordinates of each pixel point in the portrait area can be obtained, and the pixel coordinates and depth value of each pixel point in the portrait area can be input into the initial projection grid to obtain the mapped pixel coordinates corresponding to each pixel point. Further, the pixel difference between the mapped pixel coordinates and the corresponding pixel coordinates of each pixel point is calculated; the average of the pixel difference values corresponding to all the pixels in the face area is calculated to obtain the first corrected size value; the average of the pixel difference values corresponding to all the pixels in the body area is calculated to obtain the second corrected size value.

The image processing method of the embodiment of the present application obtains a captured image, identifies the subject person in the captured image, and if the subject person is identified, determines the degree of distortion of the subject person, and determines whether to perform de-distortion processing on the portrait area in the captured image according to the degree of distortion of the subject person. Thus, by performing de-distortion processing only on the distorted subject person in the captured image, the original state of the captured image is retained to the greatest extent, the amount of image processing calculation is reduced, and the efficiency of de-distortion processing of the captured image is improved.

On the basis of the above embodiment, after the subject person in the captured image is identified in the above step 102, it can be determined whether to perform dedistortion processing on the portrait area in the captured image according to the proportion of the front face in the portrait area. The above process is described in detail below in conjunction with Figure 2, which is a flow chart of the second image processing method provided in the embodiment of the present application.

As shown in FIG2 , the image processing method may include the following steps:

Step 201, acquiring a captured image.

Step 202: Identify the subject person in the captured image.

In the embodiment of the present application, the implementation process of step 201 and step 202 can refer to the implementation process of step 101 and step 102 in the above embodiment, which will not be repeated here.

Step 203: If the subject person is not identified, the number of portrait areas in the captured image is queried.

In the embodiment of the present application, when identifying the subject person in the captured image, there may be a situation where the subject person is not identified. In this case, the number of portrait areas in the captured image is queried.

Step 204: determine whether the number of portrait regions is greater than or equal to a first threshold.

In the embodiment of the present application, it is determined whether to perform dedistortion processing on the portrait in the captured image by judging whether the number of portrait areas is greater than or equal to a first threshold.

Step 205: If the number of portrait regions is less than a first threshold, it is determined that there is no need to perform dedistortion processing on the portrait regions in the captured image.

In one possible case, the number of portrait areas in the captured image is small, and the number of portrait areas is less than a first threshold. In this case, it can be determined that the captured image is a landscape photo, and it is determined that there is no need to perform dedistortion processing on the portrait areas in the captured image.

It can be understood that when an electronic device captures an image, the object being photographed is a natural scenery, but there is a portrait area in the photographed image. In this case, there is no need to perform de-distortion processing on the portrait area in the photographed image.

Step 206: If the number of portrait areas is greater than or equal to the first threshold, the proportion of front faces in the faces presented in the portrait areas is counted.

The proportion of frontal faces in faces refers to the proportion of frontal faces in faces presented in the portrait area. For example, if there are 15 faces presented in the portrait area and 13 of them are frontal, the proportion of frontal faces is 13/15.

In one possible case, there are a large number of portrait areas in the captured image, and the number of portrait areas is greater than or equal to a first threshold. In this case, the captured image may be a group photo of multiple people, or it may be a random shot.

Therefore, it is necessary to further count the proportion of the front face in the faces presented in the portrait area, so as to determine the shooting scene according to the proportion of the front face in the faces, so as to determine whether to perform de-distortion processing on the portrait area in the captured image.

Step 207, determining whether the proportion of the front face is greater than a second threshold.

In an embodiment of the present application, after obtaining the proportion of frontal faces in the faces presented in the portrait area, the proportion of frontal faces is compared with a set second threshold to determine whether to perform dedistortion processing on the portrait area in the shooting scene based on the comparison result.

Step 208: If the proportion of the front face is greater than the second threshold, determine to perform dedistortion processing on the portrait area in the captured image.

In one possible scenario, the proportion of front faces in the faces presented in the portrait area is greater than the second threshold. In this case, the shooting scene can be determined as a group photo. Therefore, it is necessary to perform dedistortion processing on the portrait area in the shooting scene.

Step 209: If the proportion of the front face is less than or equal to or greater than the second threshold, it is determined that there is no need to perform dedistortion processing on the portrait area in the captured image.

In one possible scenario, the proportion of front faces in the faces presented in the portrait area is less than or equal to the second threshold. In this case, the shooting scene can be determined as a random shot, and there is no need to perform dedistortion processing on the portrait area in the captured image.

The image processing method of the embodiment of the present application obtains a captured image and identifies the main person in the captured image. If the main person is not identified, the number of portrait areas in the captured image is queried. If the number of portrait areas is greater than or equal to a first threshold, the proportion of the front face in the faces presented in the portrait area is counted. If the proportion of the front face is greater than a second threshold, it is determined that the portrait area in the captured image needs to be dedistorted; if the proportion of the front face is less than or equal to or greater than the second threshold, it is determined that the portrait area in the captured image does not need to be dedistorted. Therefore, when the main person is not identified in the captured image, whether the portrait area needs to be dedistorted is triggered based on whether the proportion of the front face in the faces presented in the portrait area is greater than the second threshold, thereby effectively avoiding scenes that do not require image correction, such as landscapes and casual shots, and retaining the original state of the image to the greatest extent.

On the basis of the above embodiment, when determining the degree of distortion of the subject in step 103, the degree of distortion of the subject may also be determined according to the size of the subject's field of view FOV. The above process is described in detail below in conjunction with FIG3 , which is a flowchart of the third image processing method provided in the embodiment of the present application.

As shown in FIG. 3 , the above step 103 may further include the following steps:

Step 301, determining the field of view (FOV) of the subject person according to the position of the subject person in the captured image.

The FOV of the subject refers to the offset angle of the subject relative to the optical axis of the camera.

In the embodiment of the present application, after the main character in the captured image is identified, the FOV of the main character is determined according to the position of the main character in the captured image.

As a possible implementation method, a face recognition model is used to perform face recognition on the captured image to identify the position of the main person in the captured image, and then a three-dimensional model of the main person mapped to the imaging image is established. According to the coordinate value of the coordinate point (x, y), the offset angle of the main person relative to the camera optical axis is calculated.

Step 302, determining whether the FOV is less than a predetermined angle threshold.

The angle threshold is a preset angle value. For example, the angle threshold can be set to 60°.

In the embodiment of the present application, after the FOV of each subject character is determined based on the position of each subject character in the captured image obtained by identifying the captured image, it is determined whether the FOV of the subject character is less than the angle threshold.

Step 303: If it is determined that the FOV is smaller than the angle threshold, it is determined that the subject is not distorted.

In a possible scenario, if it is determined that the FOV of the main character is smaller than the angle threshold, in this case, it is determined that the main character has no distortion, and there is no need to perform de-distortion processing on the main character.

Step 304: If it is determined that the FOV is greater than or equal to the angle threshold, the real contour is predicted based on the imaging contour of the subject.

In one possible case, if it is determined that the FOV of the subject is greater than or equal to the angle threshold, in this case, it is necessary to further determine the degree of distortion of the subject to determine whether to perform de-distortion processing on the portrait area corresponding to the subject based on the degree of distortion of the subject.

In the embodiment of the present application, it is determined that the FOV of the subject is greater than or equal to the angle threshold, and the real contour is predicted based on the imaging contour of the subject. Specifically, after the imaging contour of each subject is obtained by performing key point detection on the captured image, the real contour of the subject is calculated using the conformal projection method.

Step 305: Determine the degree of distortion of the subject person according to the difference between the imaged outline and the real outline.

In the embodiment of the present application, after predicting the real contour of the corresponding subject character according to the imaging contour of the subject character, the degree of distortion of the subject character is determined according to the difference between the imaging contour and the real contour.

In one possible case, the difference between the imaged outline of the subject and the real outline is large, and it can be determined that the distortion degree of the subject is large, for example, the overall size of the subject changes greatly, or the displacement degree of the key points of the face is large.

In another possible case, the difference between the imaged outline of the main character and the real outline is small, so it can be determined that the distortion degree of the main character is small.

The image processing method of the embodiment of the present application determines the FOV of the subject according to the position of the subject in the captured image. If the FOV is determined to be less than the angle threshold, it is determined that the subject is not distorted. If the FOV is determined to be greater than or equal to the angle threshold, the real contour is predicted according to the imaging contour of the subject; and the degree of distortion of the subject is determined according to the difference between the imaging contour and the real contour. Thus, by comparing the FOV of the subject with the angle threshold, when the FOV is greater than or equal to the angle threshold, the degree of distortion of the subject is determined according to the difference between the imaging contour of the subject and the real contour, and thus whether to perform de-distortion processing on the portrait area corresponding to the subject according to the degree of distortion of the subject is determined, and by reducing the processing of the background and reducing the number of portrait areas to be processed, the efficiency of image processing is improved.

On the basis of the above embodiment, after the portrait area in the captured image is dedistorted, the straight line segments in the captured image can also be identified, and the portrait area in the captured image can be dedistorted according to the straight line segments in the captured image, so as to keep the straight line segments in the same shape before and after dedistortion, so as to improve the accuracy of dedistortion of the portrait area. The above process is described in detail below in conjunction with FIG4, which is a flowchart of the fourth image processing method provided in the embodiment of the present application.

As shown in FIG4 , the image processing method may further include the following steps:

Step 401, identifying straight line segments in a captured image.

In the embodiment of the present application, after acquiring the captured image, straight line segments in the captured image can be further identified.

As a possible implementation method, Hough transform can be used to identify straight line segments in captured images. Among them, Hough transform is one of the basic methods for identifying geometric shapes from images in image processing. It is widely used and has many improved algorithms. Hough transform is mainly used to separate geometric shapes with certain common features (such as straight lines, circles, etc.) from images. The most basic Hough transform is to detect straight line segments from images.

It should be noted that the present application does not limit the method for identifying straight line segments in the captured image, and other straight line detection methods may also be used.

Step 402 : Dedistorting the portrait region in the captured image according to the straight line segments in the captured image, so as to keep the straight line segments in the same shape before and after dedistortion.

In an embodiment of the present application, after the straight line segments in the captured image are identified, the portrait area in the captured image can be dedistorted based on the straight line segments in the captured image to keep the straight line segments in the same shape before and after dedistortion.

It is understandable that a straight line in a three-dimensional space is still a straight line when projected onto an image plane, but due to the influence of image sensor performance, a straight line in a three-dimensional space may be a curve when projected onto a plane. Therefore, it is necessary to dedistort the portrait area in the captured image based on the straight line segments in the captured image, so that the straight line segments in the three-dimensional space are still straight line segments with consistent shapes after being projected onto a plane.

The image processing method of the embodiment of the present application, by identifying the straight line segments in the captured image, dedistorts the portrait area in the captured image according to the straight line segments in the captured image, so as to keep the straight line segments in the same shape before and after dedistortion. Thus, by dedistorting the portrait area according to the straight line segments in the captured image, it is ensured that the straight line segments have the same shape before and after dedistortion, thereby retaining the original state of the captured image to the greatest extent.

As a possible implementation, in the above step 401, multiple edge points can also be determined from each pixel point according to the gradient value and pixel value of each pixel point in the captured image, multiple edge points are fitted to obtain each initial straight line segment, and multiple initial straight line segments are merged to obtain the straight line segment in the captured image. The above process is described in detail below in conjunction with FIG5, which is a flowchart of the fifth image processing method provided in an embodiment of the present application.

As shown in FIG. 5 , the above step 401 may further include the following steps:

Step 501 : determining a plurality of edge points from each pixel according to the gradient value of each pixel in the captured image and the pixel values of adjacent pixels.

In the embodiment of the present application, the gradient value of each pixel in the captured image includes the gradient and gradient direction of each pixel. When there is an edge in the image, there must be a larger gradient value. On the contrary, when there is a relatively smooth part in the image, the gray value changes less, and the corresponding gradient is also smaller. In image processing, the modulus of the gradient is referred to as the gradient, and the image composed of the image gradient is called a gradient image. The gradient of an image is equivalent to the difference between two adjacent pixels. The gradient direction of a point in the image is calculated by calculating the gradient angle between the point and its 8 neighboring points. The maximum gradient angle is the gradient direction. Among them, the 8 neighboring points are 8 points of the top, bottom, left, right, upper left, upper right, lower left, and lower right of a point.

The gradient, gradient angle and gradient direction of the pixel point are described in detail below in conjunction with FIG6. The gradient, gradient angle and gradient direction of each pixel point in the detection image can be specifically explained by the Sobel operator. Among them, the Sobel operator is one of the most important operators in pixel image edge detection, and plays a vital role in information technology fields such as machine learning, digital media, and computer vision. Technically, it is a discrete first-order difference operator used to calculate the approximate value of the first-order gradient of the image brightness function. Using this operator at any point in the image will generate the gradient vector or its normal vector corresponding to the point.

As shown in Figure 6, for pixel A, the Sobel operator is used to first calculate G _x , G _y , and then the gradient angle θ = arctan (G _y /G _x ) of pixel A is calculated. The gradient direction is the direction in which the grayscale in the detected image increases, where the gradient angle of the gradient direction is greater than the gradient angle of the flat area. As shown in Figure 6, the direction in which the grayscale value increases has a large gradient angle, and the gradient direction of pixel A is the direction in which the gradient angle between pixel A and its 8 neighboring points is the largest.

In an embodiment of the present application, after acquiring the captured image, edge detection can be performed on the captured image to determine multiple edge points. The edge detection algorithm is mainly based on the first-order and second-order derivatives of the image intensity, but the derivatives are usually very sensitive to noise, so the detection image must first be filtered to remove the noise in the captured image. Among them, a common filtering method is Gaussian filtering, that is, a set of normalized Gaussian kernels are generated using a discretized Gaussian function, and then a weighted sum is performed on each point of the image grayscale matrix based on the Gaussian kernel function. The radius of the Gaussian kernel when Gaussian filtering is performed on the detection image can be adjusted according to the size of the detection image. For example, the Gaussian kernel radius can be set to 5.

Among them, Gaussian filtering is a linear smoothing filter, which is suitable for eliminating Gaussian noise and is widely used in the noise reduction process of image processing. Gaussian filtering convolves each pixel of the image with a Gaussian kernel to obtain the value of each pixel. In the convolution process, the values of the surrounding pixels are used and the distance is used as the weight to calculate the pixel at the center of the convolution kernel. The specific operation of Gaussian filtering is: use a template (or convolution, mask) of size 2*N+1 to scan each pixel in the image, and replace the value of the central pixel of the template with the weighted average gray value of the pixels in the neighborhood determined by the template.

Therefore, by performing Gaussian filtering on the captured image, the technical problem that the noise of the image affects the gradient direction of each pixel point and then affects the accuracy of straight line segment detection is avoided, thereby improving the detection accuracy of the straight line segment.

The edge detection method in the embodiments of the present application includes but is not limited to the canny edge detection method, the prewitt edge detection method, and the like.

As a possible implementation method, after determining the gradient value of each pixel point and the pixel value of the adjacent pixel point in the captured image, for each pixel point, the gradient value of each pixel point is compared with a first gradient threshold. In one possible case, if the gradient value of a certain pixel point is greater than the first gradient threshold, the first adjacent pixel point adjacent to the corresponding pixel point in the gradient direction is queried; if the difference between the pixel value of the corresponding pixel point and the first adjacent pixel point is greater than the second gradient threshold, the corresponding pixel point is determined to be an edge point.

As an example, taking the first adjacent pixel points as the pixel points in the 8-neighborhood of each pixel point as an example, for each pixel point in the detection image, if the gradient value is greater than the first gradient threshold, the gradient value of the corresponding pixel point is calculated differentially with the gradient value of the pixel points in the 8-neighborhood. If the difference in its gradient direction with the gradient value of the pixel points in the 8-neighborhood is greater than the second gradient threshold, the corresponding pixel point is determined to be an edge point.

It should be noted that when determining multiple edge points from each pixel point of the enhanced captured image, some noise points may be determined as edge points. Therefore, it is necessary to further screen each edge point to filter out the noise points in the image, which is beneficial to improve the accuracy of straight line segment detection.

In the embodiment of the present application, after determining multiple edge points from each pixel point, for each edge point, the second adjacent pixel point adjacent to the corresponding edge point in the gradient direction is queried, and if the difference between the gradient value of the corresponding edge point and the second adjacent pixel point is greater than the third gradient threshold, the corresponding edge point is retained, and if the difference between the gradient value of the corresponding edge point and the second adjacent pixel point is less than or equal to the third gradient threshold, the corresponding edge point is filtered out. Thus, by filtering the edge points, the noise points in the image are filtered out, which is conducive to improving the recognition rate of the straight line detection method.

Step 502: fit multiple edge points to obtain multiple initial straight line segments; wherein each initial straight line segment is obtained by fitting edge points with similar gradient directions.

In an embodiment of the present application, after multiple edge points are determined from each pixel based on the gradient value of each pixel in the captured image and the pixel values of adjacent pixels, since the multiple edge points are multiple discrete points, the multiple edge points need to be fitted to obtain multiple initial straight line segments.

It should be noted that each initial straight line segment can be obtained by fitting edge points with similar gradient directions. Specifically, after determining multiple edge points based on the gradient value of each pixel point in the captured image and the pixel value of the adjacent pixel point, the edge points with similar gradient directions among the multiple edge points are determined as a set. Furthermore, the multiple edge points can be divided into multiple sets. Among them, the edge points in the same set have similar gradient directions. For each set, the edge points in the corresponding set are fitted to obtain each initial straight line segment.

Step 503: merge multiple initial straight line segments to obtain straight line segments in the captured image.

In the embodiment of the present application, due to the influence of noise in the image, edge segments in the captured image may be cut off, resulting in discontinuity of the image edge. Therefore, it is necessary to merge multiple initial straight line segments obtained by fitting multiple edge points to obtain the target straight line segment in the captured image.

The image processing method of the embodiment of the present application determines multiple edge points from each pixel point according to the gradient value of each pixel point in the captured image and the pixel value of the adjacent pixel point, fits the multiple edge points, and obtains multiple initial straight line segments; wherein each initial straight line segment is obtained by fitting edge points with similar gradient directions, and multiple initial straight line segments are merged to obtain the target straight line segment in the detected image. The method fits multiple edge points determined in each pixel point in the captured image to obtain each initial straight line segment, and then merges the multiple initial straight line segments to obtain the target straight line segment in the detected image. Since it is not necessary to repeatedly process each pixel point in the image, the straight line segment in the image can be detected quickly, thereby improving the speed of straight line detection in the image.

On the basis of the above embodiment, in the above step 502, when fitting multiple edge points to obtain multiple initial straight line segments, multiple sets can also be determined first according to edge points with similar gradient directions among the multiple edge points, and then, for each set, the edge points in the corresponding set are fitted to obtain an initial straight line segment. The above process is described in detail below in conjunction with FIG. 7, which is a flow chart of the sixth image processing method provided in the embodiment of the present application.

As shown in FIG. 7 , the above step 502 may further include the following steps:

Step 601, determining multiple sets according to edge points with similar gradient directions among multiple edge points; wherein the edge points in the same set have similar gradient directions.

In an embodiment of the present application, after multiple edge points are determined from each pixel based on the gradient value of each pixel in the captured image and the pixel values of adjacent pixels, edge points with similar edge point gradient directions are divided into the same set to obtain multiple sets.

As a possible implementation method, for multiple edge points, an edge point is determined as the initial reference point from the edge points that have not been added to any set, and the edge point whose gradient direction difference between the query and the reference point is less than an angle threshold and is adjacent to the reference point, the queried edge point and the reference point are added to the same set.

In the embodiment of the present application, after determining multiple edge points from each pixel point, each edge point can be sorted according to the size of the gradient value of each edge point, and among the edge points not added to any set, the edge point with the largest gradient can be used as the initial reference point. The edge points adjacent to the reference point can be edge points within the initial reference point 8 neighborhoods, that is, the 8 points above, below, left, right, upper left, upper right, lower left, and lower right of the reference point.

For example, the difference between the gradient angle of the reference point and the gradient direction of each edge point in the 8-neighborhood can be calculated. Assuming that the gradient direction difference between the edge points above and to the upper left of the reference point and the reference point is less than the angle threshold, the edge points above and to the upper left can be added to the same set together with the reference point.

In an embodiment of the present application, if the degree of discreteness of the gradient direction of each edge point in the same set is less than or equal to the set discreteness, the queried edge point is used as the updated reference point to repeatedly execute the step of adding the queried edge point and the reference point to the corresponding set for the edge point whose gradient direction difference between the query and the reference point is less than the angle threshold and is adjacent to the reference point, until the degree of discreteness of the gradient direction of each edge point in the corresponding set is greater than the set discreteness.

Step 602: For each set, edge points in the corresponding set are fitted to obtain an initial straight line segment.

In the embodiment of the present application, after determining multiple sets based on edge points with similar gradient directions among multiple edge points, multiple edge points in each set are fitted to obtain an initial straight line segment.

In the embodiment of the present application, fitting is performed on multiple edge points in each set, that is, connecting the multiple edge points in each set with a straight line segment to obtain an initial straight line segment.

The image processing method of the embodiment of the present application determines multiple sets based on edge points with similar gradient directions among multiple edge points; wherein the edge points in the same set have similar gradient directions, and for each set, the edge points in the corresponding set are fitted to obtain an initial straight line segment. Thus, by fitting the edge points in each set, the initial straight line segment of the corresponding set is obtained, thereby realizing the operation of combining discrete edge points.

In order to implement the above embodiment, the present application also proposes an image processing device.

FIG8 is a schematic diagram of the structure of an image processing device provided in an embodiment of the present application.

As shown in FIG. 8 , the image processing device 700 may include: an acquisition module 710 , a recognition module 720 , a determination module 730 and a processing module 740 .

The acquisition module 710 is used to acquire the captured image.

The recognition module 720 is used to recognize the subject person in the captured image.

The determination module 730 is used to determine the degree of distortion of the subject person if the subject person is identified.

The processing module 740 is used to determine whether to perform de-distortion processing on the portrait area in the captured image according to the distortion degree of the subject.

As a possible scenario, the image processing device 700 may further include:

The query module is used to query the number of portrait areas in the captured image if the subject person is not recognized.

The statistical module is used to count the proportion of front faces among the faces presented in the portrait areas when the number of portrait areas is greater than or equal to the first threshold.

The determination module 730 is further used to determine to perform dedistortion processing on the portrait area in the captured image if the proportion of the front face is greater than the second threshold.

The determination module 730 is further configured to determine that it is not necessary to perform dedistortion processing on the portrait area in the captured image if the proportion of the front face is less than or equal to or greater than a second threshold.

As another possible situation, the determination module 730 may also be used to:

If the number of portrait regions is less than the first threshold, it is determined that there is no need to perform dedistortion processing on the portrait regions in the captured image.

As another possible situation, the identification module 720 may also be used for:

Perform face recognition on the captured image; determine the face size, face rotation angle and face clarity for each face; and regard the face whose face size, face rotation angle and face clarity meet the set conditions as the main person in the captured image.

As another possible situation, the determination module 730 may also be used to:

Determine the FOV of the subject according to the position of the subject in the captured image;

If it is determined that the FOV is less than the predetermined angle threshold, it is determined that the subject is not distorted;

If it is determined that the FOV is greater than or equal to the angle threshold, the real contour is predicted based on the imaging contour of the subject;

The degree of distortion of the subject is determined based on the difference between the imaged outline and the real outline.

As another possible situation, the image processing device 700 may further include:

The line segment recognition module is used to recognize straight line segments in the captured image.

The dedistortion module is used to dedistort the portrait area in the captured image according to the straight line segments in the captured image, so as to keep the straight line segments in the same shape before and after the dedistortion.

As another possible case, the line segment recognition module can also be used to:

The determination unit is used to determine a plurality of edge points from each pixel point according to the gradient value of each pixel point in the captured image and the pixel values of adjacent pixels.

The fitting unit is used to fit multiple edge points to obtain multiple initial straight line segments; wherein each initial straight line segment is obtained by fitting edge points with similar gradient directions.

The merging unit is used to merge multiple initial straight line segments to obtain straight line segments in the captured image.

As another possibility, the fitting unit can also be used to:

According to edge points with similar gradient directions among multiple edge points, multiple sets are determined; wherein the edge points in the same set have similar gradient directions; for each set, the edge points in the corresponding set are fitted to obtain an initial straight line segment.

As another possibility, the fitting unit can also be used to:

Determine the initial reference point from the edge points that have not been added to either set;

The gradient direction difference between the query and the reference point is less than the angle threshold, and the edge point is adjacent to the reference point;

Add the queried edge points and reference points to the same set;

If the degree of discreteness of the gradient direction of each edge point in the same set is less than or equal to the set discreteness, the queried edge point is used as the updated reference point to repeat the step of adding the queried edge point and the reference point to the corresponding set for the edge point whose gradient direction difference between the query and the reference point is less than the angle threshold and is adjacent to the reference point, until the degree of discreteness of the gradient angle of each edge point in the corresponding set is greater than the set discreteness.

It should be noted that the above explanation of the embodiment of the image processing method is also applicable to the image processing device of this embodiment, and will not be repeated here.

The image processing device of the embodiment of the present application obtains a captured image, identifies the subject person in the captured image, and if the subject person is identified, determines the degree of distortion of the subject person, and determines whether to perform de-distortion processing on the portrait area in the captured image according to the degree of distortion of the subject person. Thus, by performing de-distortion processing only on the distorted subject person in the captured image, the original state of the captured image is retained to the greatest extent, the amount of image processing calculation is reduced, and the efficiency of de-distortion processing of the captured image is improved.

In order to implement the above embodiments, the present application also proposes an electronic device. Figure 9 is a structural diagram of an electronic device according to an embodiment of the present application. As shown in Figure 9, the electronic device 110 includes a memory 111, a processor 112, and a computer program stored in the memory 111 and executable on the processor 112. When the processor executes the program, the image processing method described in the above embodiments is implemented.

In order to implement the above embodiments, the present application also proposes a non-temporary computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the image processing method described in the above embodiments is implemented.

In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.

In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In the description of this application, the meaning of "plurality" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, fragment or portion of code comprising one or more executable instructions for implementing the steps of a custom logical function or process, and the scope of the preferred embodiments of the present application includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in reverse order depending on the functions involved, which should be understood by technicians in the technical field to which the embodiments of the present application belong.

The logic and/or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute the instructions), or in combination with these instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in combination with these instruction execution systems, devices or apparatuses. More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk box (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting or processing in other suitable ways if necessary, and then stored in a computer memory.

It should be understood that the various parts of the present application can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

A person skilled in the art may understand that all or part of the steps in the method for implementing the above-mentioned embodiment may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, which, when executed, includes one or a combination of the steps of the method embodiment.

In addition, each functional unit in each embodiment of the present application may be integrated into a processing module, or each unit may exist physically separately, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.

The storage medium mentioned above may be a read-only memory, a disk or an optical disk, etc. Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limiting the present application. A person of ordinary skill in the art may change, modify, replace and modify the above embodiments within the scope of the present application.

Claims (12)

1. An image processing method, characterized in that the method comprises: Acquire captured images; Identifying a subject person in the captured image; If the subject person is identified, determining the degree of distortion of the subject person; According to the degree of distortion of the subject person, determine whether to perform de-distortion processing on the portrait area in the captured image, wherein, when determining to perform de-distortion processing on the portrait area, the portrait area is divided into a face area and a body area, and the portrait area is corrected and calculated according to a preset initial projection grid to obtain a first correction size value corresponding to the face area and a second correction size value corresponding to the body area, and among the first correction size value and the second correction size value, a target correction size value that meets a preset condition is determined, so as to perform correction processing on the portrait area according to the target correction size value to obtain a de-distorted captured image. 2. The image processing method according to claim 1, characterized in that after identifying the subject person in the captured image, it also includes: If the subject person is not identified, querying the number of the portrait areas in the captured image; If the number of the portrait areas is greater than or equal to a first threshold, the proportion of front faces in the faces presented in the portrait areas is counted; If the proportion of the front face is greater than a second threshold, determining to perform a dedistortion process on the portrait area in the captured image; If the proportion of the front face is less than or equal to or greater than the second threshold, it is determined that there is no need to perform dedistortion processing on the portrait area in the captured image. 3. The image processing method according to claim 2, characterized in that after querying the number of the portrait areas in the captured image, it also includes: If the number of the portrait areas is less than the first threshold, it is determined that there is no need to perform dedistortion processing on the portrait areas in the captured image. 4. The image processing method according to any one of claims 1 to 3, characterized in that the identifying the subject person in the captured image comprises: Performing face recognition on the captured image; Determine the face size, face rotation angle and face clarity for each face; The face whose face size, face rotation angle and face clarity all meet the set conditions is taken as the main person in the captured image. 5. The image processing method according to any one of claims 1 to 3, characterized in that determining the degree of distortion of the subject person comprises: Determining the FOV of the subject person according to the position of the subject person in the captured image; If it is determined that the FOV is less than the predetermined angle threshold, it is determined that the subject person has no distortion; If it is determined that the FOV is greater than or equal to the angle threshold, predicting the real contour according to the imaging contour of the subject person; The degree of distortion of the subject person is determined according to the difference between the imaging contour and the real contour. 6. The image processing method according to any one of claims 1 to 3, characterized in that after determining to perform dedistortion processing on the portrait area in the captured image, it further comprises: Identifying straight line segments in the captured image; The portrait region in the captured image is dedistorted according to the straight line segments in the captured image, so as to keep the straight line segments in the same shape before and after the dedistortion. 7. The image processing method according to claim 6, wherein the identifying straight line segments in the captured image comprises: Determining a plurality of edge points from each pixel according to a gradient value of each pixel in the captured image and pixel values of adjacent pixels; Fitting the multiple edge points to obtain multiple initial straight line segments; wherein each initial straight line segment is obtained by fitting edge points with similar gradient directions; The multiple initial straight line segments are merged to obtain straight line segments in the captured image. 8. The image processing method according to claim 7, wherein fitting the plurality of edge points to obtain a plurality of initial straight line segments comprises: Determine multiple sets according to edge points with similar gradient directions among the multiple edge points; wherein the edge points in the same set have similar gradient directions; For each set, the edge points in the corresponding set are fitted to obtain an initial straight line segment. 9. The image processing method according to claim 8, characterized in that the step of determining a plurality of sets based on edge points with similar gradient directions among the plurality of edge points comprises: Determine the initial reference point from the edge points that have not been added to either set; Querying edge points whose gradient direction difference with the reference point is less than an angle threshold and which are adjacent to the reference point; Adding the queried edge points and the reference points to the same set; If the degree of discreteness of the gradient direction of each edge point in the same set is less than or equal to the set discreteness, the queried edge point is used as the updated reference point to repeatedly execute the step of querying the edge point whose gradient direction difference with the reference point is less than the angle threshold and is adjacent to the reference point, and adding the queried edge point and the reference point to the corresponding set until the degree of discreteness of the gradient direction of each edge point in the corresponding set is greater than the set discreteness. 10. An image processing device, characterized in that the device comprises: An acquisition module, used for acquiring captured images; A recognition module, used to recognize the subject person in the captured image; A determination module, configured to determine the degree of distortion of the subject person if the subject person is identified; A processing module is used to determine whether to perform de-distortion processing on the portrait area in the captured image according to the degree of distortion of the subject person, wherein when it is determined that the portrait area is to be de-distorted, the portrait area is divided into a face area and a body area, and the portrait area is corrected and calculated according to a preset initial projection grid to obtain a first correction size value corresponding to the face area and a second correction size value corresponding to the body area, and a target correction size value that meets a preset condition is determined from the first correction size value and the second correction size value, so as to perform correction processing on the portrait area according to the target correction size value to obtain a de-distorted captured image. 11. An electronic device, characterized in that it comprises a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the image processing method according to any one of claims 1 to 9 is implemented. 12. A non-transitory computer-readable storage medium having a computer program stored thereon, wherein when the program is executed by a processor, the image processing method according to any one of claims 1 to 9 is implemented.