JP2007149092A - Multiple resolution adaptive filtering - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a system and a method for analyzing an image and extracting features of the image therefrom for use in filtering. <P>SOLUTION: Based on the features and structures, embodiments determine how to filter at different orientations and with different filter configurations. Filters utilized according to the embodiments are adaptive with respect to spatial and/or temporal aspects of the features. Image processing according to the embodiments is performed on sub-images at various levels of resolutions. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

CROSS REFERENCE TO RELATED APPLICATIONS This application is a co-pending US Provisional Application No. 60 / 739,871, filed Nov. 23, 2005, the disclosure of which is incorporated herein by reference, entitled “Multi-Resolution”. It claims the priority of “Adaptive Filtering”.

  The present invention relates generally to image processing, and more specifically to image processing that reduces image noise.

  Speckle noise includes, for example, coherent noise generated from an ultrasound image. For example, when forming an ultrasound image, a beam forming process, which is a coherent process, is typically performed to form an ultrasound beam, and an ultrasound image is derived from the ultrasound beam. Many beam forming processes result in some sort of “sesame salt noise” that is superimposed on the true image information. This noise is called “speckle noise”. A similar phenomenon occurs in radar.

  Many people in the ultrasound industry have attempted to filter out speckle noise based on synthesizing compounding techniques. That is, many people have attempted to reduce speckle noise by processing images in different frequency bands and integrating these different frequency bands together to reduce speckle noise.

  Another form of reducing speckle noise that has been attempted is the use of spatial compounding. In spatial compounding, multiple images are generated from different “look directions” or different angles of view, and these images are integrated together so that speckle noise is averaged. Excluded.

Both of the aforementioned speckle noise reduction techniques often achieve some level of speckle noise suppression. However, these techniques are not without their disadvantages. For example, frequency compounding has a certain axial resolution compromise. This is because the frequency band is divided into smaller and narrower bandwidth signals. This narrower banding results in a compromise in axial resolution. The use of spatial compounding is achieved by acquiring multiple views from different look directions, but reduces the frame rate of the final image acquisition. Thus, real-time images may have low quality motion or animation.
US Provisional Application No. 60 / 739,871 US Pat. No. 5,722,314

Summary of invention

  The present invention is directed to a system and method that reduces speckle noise by analyzing an image, extracting local features of the image, and applying an adaptive filter to such features.

  Based on various features among the features identified in the image, the embodiments provide filtering in different orientations and / or different filters to enhance image quality by effectively suppressing speckle noise. Determine the filter configuration for applying filtering using parameters. The applied filter is preferably adaptive with respect to the particular feature to which the filter is applied, for example spatially and / or temporally.

  Embodiments of the present invention perform processing on sub-images at various levels of resolution using the adaptive filters described above. For example, one high resolution image can be decomposed into multiple image representations, each having a lower resolution than the next image representation. Embodiments operate to identify local features in each such image representation and apply a filter to it, where the applied filter corresponds to the corresponding feature present in that particular image representation. Is selected using orientation and / or parameters. Information about the features in the image representation of the image can be shared between processes that apply filtering to different ones of the image representation. After filtering is applied to each image representation, the preferred embodiment reconstructs the filtered image from the filtered image representation. The above-described image decomposition, filtering of the decomposed image representation, and reconstruction of the filtered image representation may be performed multiple times on the same image (eg, iteratively or when changing or applying changes to an image) ) Can be performed.

  Various knowledge bases can be utilized in applying the adaptive filter of embodiments of the present invention. For example, a knowledge base that associates various filter parameters with aspects of a feature (eg, step function, edge, surface slope, texture, pixel intensity gradient, etc.) is used for a particular feature identified in the image It can be utilized in selecting an adaptive filter and / or adaptive filter parameters. In addition or alternatively, a knowledge base has been identified in the image that associates various filter parameters with features that are typically present in a particular image type (eg, a particular anatomy, a particular procedure, etc.) It can be utilized in selecting an adaptive filter and / or adaptive filter parameters to be used for a particular feature.

  The foregoing has outlined rather broadly the features and technical advantages of the present invention in order that the detailed description of the invention that follows may be better understood. Additional features and advantages of the invention will be described hereinafter which form the subject of the claims of the invention. Those skilled in the art should appreciate that the disclosed concepts and specific embodiments can be readily utilized as a basis for other structural modifications or designs to perform the same purposes as the present invention. Those skilled in the art will also recognize that such equivalent constructions do not depart from the spirit and scope of the invention as set forth in the appended claims. The novel features of both the organization and method of operation that appear to be unique to the invention, together with further objects and advantages, will be better understood from the following description when considered in conjunction with the accompanying drawings. However, it should be particularly understood that each of the drawings is provided for purposes of illustration and description only and is not intended as a definition of the limitations of the invention.

  For a more complete understanding of the present invention, reference will now be made to the following description, taken in conjunction with the accompanying drawings.

  Turning attention to FIG. 1, a representation of applying an adaptive filter to an image 100 according to an embodiment of the present invention is shown. Specifically, FIG. 1 includes circles 111-114 and ellipses 121-127 representing exemplary filters according to an embodiment of the present invention. These circles and ellipses are overlaid on the image 100, such as with respect to various corresponding features present in the image (eg, structure, texture, gradient, slope, function, etc.). In operation according to a preferred embodiment of the present invention, the aforementioned filters are of different sizes, applied in different orientations with respect to the image, and / or use various parameters, but remove speckle noise by smoothing, and Developed and / or applied to images to filter out. Such a filter is preferably adapted to local features in order to provide the best quality of filtering performance. For example, the filtering processors of embodiments average the pixels at different locations by using the filters described above to avoid averaging different pixels while averaging similar pixels. Thus, the system preferably adaptively determines whether it is desirable to include a pixel in the filtering process. One or more knowledge bases can be utilized in selecting filters and / or filter parameters for use with respect to particular features of the image, such as image 100.

  Embodiments of the present invention perform sub-image processing with respect to providing image filtering. For example, the multi-resolution sub-image processing step of the embodiments decomposes the image into sub-images or image representations of different resolution, where one or more adaptive filters are used to suppress each speckle noise. Applied to various features present in the image. The filters used for any such sub-image can be of different sizes, applied in different orientations with respect to the image, and / or use various parameters, as described above. That is, embodiments of the present invention implement multiple filters that include behavior that is locally dependent on features within one of the sub-images. In operation according to a preferred embodiment, if a sub-image is processed using the adaptive filter described above, the processed sub-images are combined to reconstruct the image, which is then Presented to the user or otherwise used or stored as a filtered image.

  Examples of adaptive filter characteristics according to various embodiments include flat surface filtering and smoothing, sloping surface filtering, preserving sharp edges between surfaces, noise filtering on ridge features, sharp corners Can include storage. For example, a filter that can be implemented with respect to a particular feature having a flat surface can average the pixels on that same surface, i.e., pixels having similar characteristics. However, if the surface is not a flat surface but a curved surface with a certain slope associated therewith, the embodiments group and filter pixels along the same slope without disturbing the slope. A filter can be implemented. Filters implemented in accordance with embodiments preserve the sharpness of edges when there are sharp edges between surfaces, such as top and side surfaces (eg, a cubic view when multiple surfaces are visible). In order to minimize edge breakage and distortion. When the image feature or structure includes a ridge (eg, a structure resembling a line), filters implemented in accordance with embodiments of the present invention preserve the shape of the ridge and preserve the sharpness of the ridge. When there is a line intersection (eg, two ridges intersect) and an angle is produced at the position of the intersection, the filter implemented in accordance with embodiments of the present invention preserves that angle to be very sharp To do. In the operation of the preferred embodiment of the present invention, one or more filters are developed using the aforementioned characteristics associated with the identified features in the image, which results in high image quality in the speckle reduction process. Becomes easier.

  Turning attention to FIGS. 2A, 2B, and 3, it can be seen that various embodiments of the present invention can be applied in multiple dimensions. For example, an algorithm implementing the concept of the present invention can be applied to one-dimensional signal processing, two-dimensional image processing, three-dimensional image processing, and four-dimensional image processing. Four-dimensional signal processing according to embodiments refers to three-dimensional space and time (eg, X, Y, Z, and time). Three-dimensional signal processing according to embodiments refers to two-dimensional space and time (eg, X, Y, and time) or three-dimensional space (eg, X, Y, and Z).

  An example of two-dimensional signal processing is shown for the image sequence of FIG. 2A, where the t-axis is time and the X-axis and Y-axis are the spatial dimensions of the two-dimensional image. In the example shown in FIG. 2A, the image processing that results in the filtering described above is provided independently for each frame in the time sequence, shown as frames 201-209 in this figure (ie, intra-frame processing). That is, this process is performed on a frame-by-frame basis without using interframe information (eg, information obtained from the next frame) available from the time sequence of frames. Various embodiments of the present invention may use one or both of inter-frame (spatial information in this example) filter feature and intra-frame (in this example temporal information) filter feature.

  An example of three-dimensional signal processing is shown in FIG. 2B. Specifically, the example shown in FIG. 2B illustrates the use of interframe information. In the illustrated interframe signal processing, consecutive frames, such as the frames in groups 211-214, are included in this processing to improve filter performance. For example, a filter applied with respect to features present in successive frames of group 211 in FIG. 2B can be oriented along the t-axis to utilize information available in successive frames. Such a frame group can be defined by a series of frames whose features are correlated, by any number of frames, by a moving window, etc. Embodiments of the present invention can implement frame groups that include different frames (eg, group 211 for the first feature and group 212 for the second feature) to provide filtering for various image features. . However, the use of such interframe information for various features is selected as described above, and the use of such interframe information for various features can be utilized to provide improved image processing when relevant information is provided in interframes. it can.

  It should be appreciated that the above concepts can be applied with respect to different degrees of spatial information and are therefore not limited to use with two-dimensional spatial image processing. The example of FIG. 3 shows the application of the above concept for a three-dimensional object sequence. For example, if the image includes a three-dimensional block of data, but this block of data is acquired at a different time, a block image sequence (shown as blocks 310 and 302 in this figure) can be acquired, as described above. The concept can be applied to improve image quality.

  The following equation provides a filter formulation describing two categories of filters according to various embodiments.

上に示されたフィルタの第１カテゴリ（式（１））は、対称フィルタを含む。上に示されたフィルタの第２カテゴリ（式（２））は、非対称フィルタを含み、ここで、ｕは、特徴の支配的方位（ｄｏｍｉｎａｎｔ ｏｒｉｅｎｔａｔｉｏｎ）である。対称フィルタと非対称フィルタの両方を、適応式とすることができる。この例示的フィルタ定式化は、複数の異なるパラメータの関数である。ここで、ｒ、ｘ、ｙ、およびｚは、空間情報を含み、ｔは、時間情報を含み、ｇは、グレイスケール情報（たとえば、差分グレイスケール情報）を含み、ｆは、総称項（たとえば、他の関連情報）である。

The first category of filters shown above (Equation (1)) includes symmetric filters. The second category of filters shown above (Equation (2)) includes asymmetric filters, where u is the dominant orientation of the feature. Both symmetric and asymmetric filters can be adaptive. This exemplary filter formulation is a function of several different parameters. Where r, x, y, and z include spatial information, t includes temporal information, g includes grayscale information (eg, differential grayscale information), and f is a generic term (eg, , Other related information).

  The coordinate system in which the speckle reduction operation is performed according to embodiments of the present invention can be a polar coordinate system (eg, using a radial coordinate such as a radius or a distance r) or a Cartesian coordinate system (eg, an X-axis, a Y-axis, and a Z-axis). It should be understood that the coordinate along the axis can be used). The ultrasound information is basically data acquired from different look directions, so that this data can be assembled into an image according to polar coordinates. However, since linear arrays are usually rectangular, most display modes are Cartesian. Scan heads typically use polar coordinates, and displays typically use Cartesian coordinates, so often from the polar coordinate system used for image data acquired by an ultrasound system scan head, the ultrasound system There is a reconstruction process to the Cartesian coordinate system used for the image data displayed by the display. That is, most ultrasound systems convert from polar coordinates to Cartesian coordinates (referred to as a scan conversion process). The filtering described herein can be applied in polar coordinate space and / or orthogonal coordinate space.

What is described on the right side of equation (1) is that a cascade (eg, several different Gaussian filters) is used. Gaussian filters are cascaded to a single filter, called a G _s. The filter equation shown is symmetric with respect to r, i.e. distance, and is therefore symmetric and uniform in all dimensions. The delta g (Δg) of the embodiment of Equation (1) includes gradient information such as gradients for gray scales in different dimensions. The f function is provided in this exemplary embodiment, for example, to deal with generic functions. It should be appreciated that a plurality of such f-functions can be used in accordance with embodiments of the present invention, such as where there are a plurality of different additional relevant information.

  Equation (2) provides an asymmetric filter, which is not uniform in all dimensions. That is, the filter applies filtering differently with respect to the selected orientation, such as along the feature axis as opposed to an axis orthogonal to the feature axis. The use of such asymmetric filters according to embodiments of the present invention may be particularly useful in preserving aspects of features in the filtered image. Similar to equation (1) above, equation (2) shown above uses a cascade (eg, multiple different Gaussian filters). It should be appreciated that this exemplary asymmetric filter can be resolved in different orientations. The filter orientation can be determined by analyzing local features to extract one or more feature orientations. The filter can then preferably be applied along the dominant orientation of the feature.

  Turning attention to FIGS. 4A and 4B, an illustrative example of applying adaptive filtering according to an embodiment of the present invention is shown. To simplify the described concept, the example of FIGS. 4A and 4B provides a diagram of one-dimensional signal processing adaptive filtering. As shown in FIG. 4A, there is a signal that is a step function, illustrated as a step function signal 401. If noise, such as that represented by the noise signal 402, is introduced into the step function signal, the step function signal becomes a noisy step function signal, such as that represented by the noisy step function signal 403. Become. It should be appreciated that the signal of FIG. 4 represents a single feature extracted from, for example, an image signal. There may be many other possible features in a given signal that are extracted based on similar trace modes.

  The adaptive filter of the embodiment of the present invention, shown as filter 410 in FIG. 4B, removes the noise, and thus the original signal (step shown in FIG. 4A, shown as filtered step function signal 404). It is preferably applied to a noisy step function signal 403 to represent a filtered signal approximating the function signal 401). The application of such a filter according to embodiments is to suppress noise and preserve step function edges. Accordingly, a sharp edge of the step function remains in the filtered step function signal 404 of the illustrated embodiment. Thus, adaptive filters according to embodiments of the present invention are particularly useful with respect to feature boundaries, such as to preserve sharp feature edges, corners, lines, etc.

  FIG. 5 illustrates the application of a conventional non-adaptive filter to the aforementioned noisy step function signal to illustrate the advantages of the adaptive filter of embodiments of the present invention. The filter of FIG. 5 does not adapt to the signal. A Gaussian kernel, illustrated as filter kernel 510 and referred to herein as G (h), is illustrated and represents a conventional non-adaptive filter. A Gaussian kernel G (h) exists at the edge of the step function. Similarly, the same Gaussian kernel G (h) exists at a few points away from the edge. Thus, by applying this Gaussian kernel and taking the average of the signal, the conventional filter provides a filtered step function signal 504 where the noise is smoothed by smoothing. In addition to being removed, the edge of the step function is also removed by smoothing and is therefore no longer sharp. Specifically, the sharpness of the edges at regions 521 and 522 is now reduced by the smoothing provided by the filter, so the ultrasound image formed from this signal has edges that appear as blurred boundaries. Should bring.

  Unlike the conventional non-adaptive filter shown in FIG. 5, the adaptive filter 410 of FIG. 4B preserves the sharp edges of the step function features. FIG. 6 shows how the adaptive filter 410 of the embodiment of the present invention can operate in a one-dimensional space. Specifically, FIG. 6 shows how the extracted features, in this figure the step function, are used to control the filter applied to reduce the noise.

  The filter kernels utilized at any particular point with respect to the noisy step function signal 403 are represented in FIG. 6 as filter kernels 611-615. Unlike the filter kernel 510 of FIG. 5, the filter kernel of FIG. 6 is adaptive. Thus, when the filter kernel approaches the step function, the filter kernel is adapted to correspond to the step function feature. Note that the shape of the Gaussian kernel represented by filter kernels 613 and 612 has a sharp edge on the left side, but the right side is smooth, similar to the filter kernel 510 shown in FIG. Specifically, in the illustrated embodiment, the filter kernel 613 is set to about half of this filter kernel, and only half of this Gaussian kernel is used to filter out noisy signals. Applicable. However, the filter kernel 612 is set to 0 for only about 1/3 of the filter kernel, so that about 2/3 of the Gaussian kernel is applied to filter out noisy signals. It has become. As this filter moves away from the step function edge, the filter kernel takes a shape similar to that shown in FIG. By using the adaptive filter of the illustrated embodiment at the edge of the step function, the left side of this signal is no longer averaged to the right side of this signal.

  In the operation according to the embodiment shown in FIG. 6, when approaching the step function edge, the adaptive filter automatically reduces weights or changes the filter coefficients, and the signal to the right of this step function edge is the step function edge. It will not be averaged with the signal on the left side of the edge. It should be appreciated that the filter function described above can be applied to any signal that is examined, whether or not there is a sufficiently large signal transient to include an “edge”. However, if the feature or structure is large enough to contain an edge, it is preferred that the filter weighting is manipulated so that the pixels on either side of the edge cannot be averaged together. That is, embodiments operate to look for an edge, and when an edge is found, the filter weight is adjusted towards that edge so that the filter weight at that edge is at or near zero. . It should be appreciated that such edges can be defined in a multi-dimensional space, and thus the aforementioned concept of an adaptive filter operating on edges is not limited to the one-dimensional example shown above. Furthermore, the dimension in which such an edge is defined is not limited to a spatial dimension and can therefore be defined in time.

  FIGS. 7A and 7B show a two-dimensional case where adaptive filtering is applied similar to the one-dimensional case of FIG. 6 described above. In the two-dimensional case, embodiments of the invention are performed as described above, but in two dimensions. For example, as shown in FIG. 7A, a complete Gaussian filter (shown as filter kernel 711) is a flat surface of a two-dimensional noisy step function signal (shown as noisy step function signal 703). It can be applied to the surface (eg, upper and lower plateaus). However, near the step function edge, the filter kernel preferably reduces the filter weights so that the right side only smooths the top (eg, filter kernel 712 causes the filter to approach the step function edge). (Sometimes, 0 is set to about 1/3 of this filter kernel). Therefore, the information on the left side cannot be averaged together with the information on the right side of the edge. FIG. 7B shows that the sharp edge of the step function is preserved in the filtered step function signal 704 after application of this adaptive filter.

  As noted above, equations (1) and (2) used in connection with the above-described adaptive filter embodiments are cascaded filters (in the above example, cascaded Gaussian functions), which are This results in a relatively difficult filter. However, it is preferable for the filter kernel to automatically reduce the weight when the local feature is at the edge. When the weights are reduced, the filters of the embodiments are no longer complete kernels at the edges, as shown by the filter kernels 612 to 615 of FIG. The filter can be much less demanding than the full filter kernel.

  As mentioned above, the filter categories represented by equations (1) and (2) include symmetric filters (equation (1)) and asymmetric filters (equation (2)). The difference is that the asymmetric filter has an orientation applied to it and the symmetric filter is isotropic for all different directions. Thus, a filter implemented using equation (1) provides filtering as a function of distance, and a filter implemented using equation (2) provides filtering as a function of gradient. The concepts described above can be given for both filter categories (eg, adjusting filter weights with respect to distance from features and / or with respect to feature gradients).

  8A and 8B show an exemplary symmetric adaptive filter at the ridge, according to an embodiment of the present invention. The filter according to the illustrated embodiment is adapted to the image features. As shown in FIG. 8A, a perfect Gaussian filter (shown as filter kernel 811) is applied to a flat surface (eg, background) of a noisy ridge signal (shown as noisy ridge signal 803). Can be applied. However, it is preferred that the filter kernel reduce the filter weight at the ridge so that only the top of the ridge feature is flattened (eg, the filter kernel 812 is applied to the ridge feature when the filter kernel is 0 is set to about 3/1 on the left side and about 3/1 on the right side of the kernel). Therefore, the information from the left and right background surfaces of the ridge feature cannot be averaged along with the information on the ridge. FIG. 8B shows that after applying this adaptive filter, the sharp edges of the ridge are preserved in the filtered ridge signal 804.

  9A and 9B illustrate an exemplary symmetric adaptive filter according to an embodiment of the present invention at a feature intersection (eg, a ridge intersection). The filter of the illustrated embodiment is adapted to image features near the center. As shown in FIG. 9A, a perfect Gaussian filter (shown as filter kernel 911) is applied to a flat surface (eg, background) of a noisy ridge crossing signal (shown as noisy ridge crossing signal 903). ) Can be applied. However, it is preferred that the filter reduce the filter weights at the ridge and ridge intersection (eg, the filter kernel 912 has zeros in the parts of the filter kernel corresponding to the portion of the ridge intersection to which the filter is applied. Set). The filter kernel 912 of the illustrated embodiment takes a more complex shape compared to, for example, the filter kernel 812, which preserves the ridge intersection corners in addition to preserving the ridge edges. I want to be understood. Therefore, the information from the left and right background surfaces of the ridge feature cannot be averaged along with the information on the ridge. FIG. 9B shows that after application of this adaptive filter, the filtered ridge crossing signal 904 preserves the sharp edges of the ridge as well as the sharp edges of the ridge crossing corners.

  As noted above, according to the preferred embodiment, the adaptive filter attempts to preserve the corners and also to preserve the edge sharpness associated with feature intersection. While adaptive filters generally work well in achieving the foregoing, there are some limitations in certain applications. Specifically, using a symmetric filter, the filtering weight is a function of local features, in which case the effective filtering kernel size is smaller when the weight is suppressed. As a result, the filtering effect near the feature edge is smaller than the surface. This phenomenon can be seen in FIG. 8A, where a full Gaussian filter is applied to the background region, but only a subset of the Gaussian filter should be applied along the ridge edge. is there. Thus, the amount of filter applied at different parts of the image should not be the same, resulting in a filtered signal that leaves more noise near the ridge compared to the region away from the ridge. FIGS. 6A and 9A illustrate this phenomenon with respect to the features therein.

  Accordingly, embodiments of the present invention implement a steerable filtering process to compensate for the aforementioned irregular filtering effects at feature edges. Embodiments deploy multiple different steerable filters according to the feature orientation. These steerable filters can be extracted or classified and the system preferably applies these filters to the current orientation to provide additional filtering around the edges. For example, a steerable filter may be provided for application along the ridge of the noisy ridge signal 803 of FIG. 8A, but a steerable filter may be provided for the background plane and ridge of the noisy ridge signal 803. Can be provided for application to the background along edges defined by intersections. Various embodiments of the present invention include steerable asymmetric filters in addition to symmetric filters to improve performance. In such an embodiment, the system applies an asymmetric filter to the feature based on the orientation of the feature.

  10A and 10B show various example steerable filters used in at least one embodiment of the present invention. According to a preferred embodiment, the orientation of local features in the filtered signal is determined (eg, the dominant angle θ of the features can be determined). An oriented steerable filter corresponding to the orientation of the local features is preferably created (eg, an asymmetric steerable filter oriented at an angle θ). This steerable filter is then applied to the feature at the determined orientation to provide a filtered image.

  Although the diagrams of FIGS. 10A and 10B show spatial centering performed in XY space, the same concept can be extended to other dimensions (eg, Z axis and time axis). For example, various embodiments can be adapted to four dimensions (eg, X, Y, Z, and time). Note that one or more of the aforementioned dimensions need not be space and can thus include time, brightness, and the like. If there are multiple dimensions (eg, three dimensions measured in XYZ), the system can have different filters with different orientations in different dimensions.

  The equation shown below represents a steerable asymmetric filter for a relatively simple two-dimensional case according to various embodiments of the invention.

ここで、図１１に示されているように、式（３）のｖは、特徴エッジに垂直な勾配方向であり、式（３）のｕは、特徴エッジに平行である。式（３）の実施形態には、上で述べた非対称フィルタ式（２）から導出された、２つの指数表現、ここでは２つのガウス式（Ｇａｕｓｓｉａｎ ｅｘｐｒｅｓｓｉｏｎ）がある。第１のガウス式は、ｕおよびｖを含み、特徴の方位を示す。次のガウス式は、ｕ方向に沿ったグレイスケールの勾配（ｇ）と、ｖ方向に沿った特徴の勾配とを表す。図１１に示された楕円は、結果のフィルタを示す表現を提供する。図示の実施形態では、このフィルタが、ｕ方向に沿って広がり、より小さい範囲まで、垂直のｖ方向に沿って広がる。これがガウス・フィルタであると仮定すると、ガウシアンの広がりは、シグマｕおよびシグマｖによって記述される。図示の実施形態のシグマｖは、シグマｕより小さい。シグマは、ＸＹ軸に関するこのｕ軸の方位であり、ＸＹ軸は、画像の品質であり、ｕは、特徴の方向である。シグマｇは、この特定のガウス関数の広がりである。

Here, as shown in FIG. 11, v in equation (3) is the gradient direction perpendicular to the feature edge, and u in equation (3) is parallel to the feature edge. In the embodiment of equation (3), there are two exponential expressions, here two Gaussian expressions, derived from the asymmetric filter equation (2) described above. The first Gaussian equation includes u and v and indicates the orientation of the feature. The following Gaussian equation represents the grayscale gradient (g) along the u direction and the feature gradient along the v direction. The ellipse shown in FIG. 11 provides a representation showing the resulting filter. In the illustrated embodiment, the filter extends along the u direction and extends along the vertical v direction to a smaller extent. Assuming this is a Gaussian filter, the Gaussian spread is described by sigma u and sigma v. The sigma v in the illustrated embodiment is smaller than the sigma u. Sigma is the orientation of this u axis with respect to the XY axis, the XY axis is the image quality, and u is the direction of the feature. The sigma g is the spread of this particular Gaussian function.

  The uv space in the illustrated embodiment is a feature space, but is a rotation transformation that can be expressed by the following equation.

したがって、ステアラブル・フィルタは、画像空間内で次の式に従って表すことができ、ここで、λ_ｖ＞＞λ_ｕかつ｜∇ｇ_ｖ｜＞＞｜∇ｇ_ｕ｜と仮定する。

Thus, a steerable filter can be expressed in image space according to the following equation, where λ _v >> λ _u and | ∇g _v | >> | ∇g _u |.

  From above, the system according to embodiments can identify the orientation of a feature and work along a particular direction of that feature using, for example, a filter kernel defined by equation (3) or (5) It should be appreciated that the filter can be adapted. This direction can be the direction of the filter itself. According to embodiments of the invention, the surface can be characterized, the gradient itself can be characterized, the position of the structure can be characterized, and so on.

  Function G is in uv space

として表すことができる。λ_ｖ＞＞λ_ｕであり、ｖが勾配方向であると仮定し、σ_ｕ＞＞σ_ｖであるものとすると、式（６）の勾配を、下で示されているように表すことができる。

Can be expressed as Assuming λ _v >> λ _u , v is the gradient direction, and σ _u >> σ _v , the gradient of equation (6) can be expressed as shown below: it can.

ここで、

here,

は、特徴エッジに平行なベクトルであり、

Is a vector parallel to the feature edge,

は、フィルタリング領域内の（ｘ，ｙ）にある点である。

Is the point at (x, y) in the filtering region.

  The gradient direction indicates that the steepness changes in a two-dimensional space. In other words, when the gray scale changes, this can be described as terrain. On the steep side, the slope is larger. Usually, it is not desirable to smooth the image in that particular direction (eg, avoid “falling off the cliff”). Accordingly, embodiments of the present invention apply a smoothing function in a direction different from the direction of the steepest gradient. For example, if the maximum gradient direction is gradient Gv, the system applies a filter along the u direction because the maximum gradient is the v direction. In other words, various embodiments apply a smoothing filter along a direction perpendicular to the direction where the gradient is maximum.

  The feature edge orientation can be found by finding eigenvectors from the Hessian matrix defined by the luminance gradient Jacobian according to embodiments of the present invention. The Hessian matrix M is represented below.

入力画像Ｉは、導関数をとる前に、まずガウス・フィルタＧによって正規化することができる（Ｊ＝Ｇ＊Ｉ）。ヘッシアン行列Ｍの固有値および固有ベクトルを、下記によって計算し、表すことができる。

The input image I can first be normalized by a Gaussian filter G (J = G * I) before taking the derivative. The eigenvalues and eigenvectors of the Hessian matrix M can be calculated and represented by:

ここで、λ_ｖ＞λ_ｕであり、ｖは、軸に対して角度θの方位にされたエッジに垂直なベクトルであり、ｕは、式（９）のエッジに平行である。

Here, λ _v > λ _u , v is a vector perpendicular to the edge oriented at an angle θ with respect to the axis, and u is parallel to the edge of Equation (9)

  Although the foregoing example utilizes eigenvectors to locate feature edges, embodiments of the present invention can implement additional or alternative techniques for locating feature edges. For example, various known digital image processing techniques, computer vision signal processing techniques, morphological image processing, etc. can be utilized in accordance with embodiments of the present invention to locate features. Embodiments of the present invention can implement, for example, fuzzy logic to locate features, where the fuzzy logic controller analyzes various attributes of the putative features to determine the best feature matching decision. It can be performed.

  FIG. 12 illustrates edge and gradient directions according to various embodiments of the invention. The image shown in FIG. 12 is a more complex two-dimensional image than that shown above with respect to FIGS. 7B, 8B, and 9B. In the embodiment of FIG. 12, the third dimension is gray scale. The gradient direction indicated by the parallel arrows indicates a certain steep change in the two-dimensional surface. The edge direction is perpendicular to the gradient direction. According to a preferred embodiment, an adaptive filter is applied along the edge direction. It should be appreciated that the filter orientation can be calculated based on the mathematical calculations of equations (6)-(9) described above.

  It should be appreciated that the object represented in FIG. 12 includes the composition of various primitive features described previously. Specifically, the object of FIG. 12 has an edge, a ridge, and a slope. When an object has such a combination of features in locality, the adaptive filter implemented by the embodiments of the present invention is a composite of various filter configurations adapted to various ones of the features. For example, it can be a combination of the filter configurations described above.

  13A-13C illustrate a simple example embodiment of a Gaussian filter kernel that can be utilized in accordance with embodiments of the present invention. FIG. 13A shows a one-dimensional Gaussian filter kernel and FIG. 13B shows a two-dimensional Gaussian filter kernel. In the illustrated example, the implementation of a Gaussian filter kernel on the continuum involves a number of points. This Gaussian filter kernel is approximated as a binomial expansion using the central limit theorem. The central limit theorem teaches that the pattern can continue until a large number of points on the Gaussian function are determined. According to this theorem, Gaussian can be approximated by a binomial kernel, and this binomial kernel can be generated by iterative averaging of the neighborhood.

FIG. 13C shows that the two-dimensional filter of FIG. 13B is steerable in at least four directions. Specifically, FIG. 13C represents a simple filter that describes a steerable filter in four directions in two dimensions. Of course, this concept can be applied in any number of directions (eg, 6 directions). The values of a, b, c, d, e, f, and g of the filter kernel of FIG. 13C can be defined according to the straightness of the Gaussian function (eg, with parameters of σ _u and σ _v ). The main direction of the embodiment shown in FIG. 13C is aa. The direction of the three points becomes bab and the rest of the directions are filled with different coefficients according to a Gaussian function and have different sigma.

  In operation according to embodiments of the present invention, the system examines features and groups information regarding filtering if the features obey certain types of criteria (eg, pixel similarity). If the criterion is satisfied, the corresponding pixel is preferably included in the filter process. Otherwise, this particular pixel is not included in the filter process. In other words, the algorithm of the present invention can operate to examine a pixel, and if that pixel is not sufficiently similar to the neighboring pixel, these pixels are not averaged together. However, if the edge orientations are very similar, embodiments can average the pixels together. In general, if a pixel is close to the surface, it is rather a flat area pixel and it is desirable to filter the pixels together. If there are sharp changes, for example because the pixels are in different regions, it is generally not desirable to filter the pixels together.

  Although the foregoing example has been described with respect to a single feature, it should be appreciated that a single image signal may include multiple features. Accordingly, embodiments of the present invention identify various features of such features, and select and / or apply one or more filters for such features as described above. To work. Further, to optimize image filtering, embodiments of the present invention perform sub-image processing with respect to providing image filtering. As noted above, embodiments of the present invention decompose an image into different resolution sub-images or image representations. One or more of the adaptive and steerable filters described above are applied to the various features present in each such sub-image. The filter used for any such sub-image can be a different size filter, applied in different orientations with respect to the image, and / or use various parameters, but each such filter Can be selected and applied as described above.

  FIG. 14 illustrates an exemplary signal path of a diagnostic ultrasound system 1400 adapted for use with various embodiments of the present invention. However, it should be appreciated that the present invention is not limited to a particular signal path. The illustrated signal path includes a scan head 1401, such as one that can include an ultrasonic transducer array as is well known in the art. Other embodiments can replace the scan head 1401 with various circuitry, such as an antenna array in a radio frequency embodiment. A front end network 1402, such as one that can be provided as a front end application specific integrated circuit (ASIC), is, for example, analog to digital signal conversion, digital to analog signal conversion, beamforming, and / or other fronts. -End processing can be provided. A signal processor 1403, such as one that can be provided as a digital signal processor (DSP), provides, for example, some level of signal filtering, synthetic aperture formation, frequency compounding, Doppler processing, and / or other advanced processing. Can do. A back end network 1405, such as that which can be provided as a back end ASIC, can provide, for example, scan conversion, video signal output, and the like. An ultrasonic signal processed by a scan head 1401, front end circuitry 1402, signal processor 1403, and back end circuitry 1405, such as a display 1406, which may include a cathode ray tube display system, a liquid crystal display system, etc. A user interface for displaying information such as a video image generated from the user to the user is provided. Additional proposed for an ultrasound system having a signal path including a scan head, front end circuitry, signal processor, back end circuitry, and display that can be adapted for use in accordance with embodiments of the present invention. Details are illustrated and described in US Pat. No. 5,722,314, the disclosure of which is incorporated herein by reference.

  In the exemplary signal path shown in FIG. 14, the adaptive filtering of an embodiment of the present invention is performed within the external DSP 1404. Specifically, the external DSP 1404 in the illustrated embodiment is interfaced with back end circuitry 1405 from which digital image information (before or after scan conversion by the back end circuitry 1405) is obtained. Received and filtered image information is provided to back-end circuitry 1405. However, alternative embodiments implement adaptive filtering in other circuits, whether inside or outside the diagnostic ultrasound system signal path and / or used in connection with a diagnostic ultrasound system be able to. For example, adaptive filtering of embodiments of the present invention can be provided as part of signal processor 1403 if desired.

  According to an embodiment of the present invention, the external DSP 1404 operates under the control of software that implements the adaptive filter kernel and steerable filter kernel described above. Specifically, the external DSP 1404 of one embodiment identifies one or more features in the digital image signal, determines the orientation of such features, and selects a filter kernel to be applied to the features And / or configure and implement an algorithm to apply a filter kernel to the image signal. If sub-image processing for image filtering is provided, the external DSP 1404 may further provide multi-resolution decomposition of the image signal and multi-resolution reconstruction of the filtered signal according to embodiments of the present invention. it can.

  Embodiments of external DSP 1404 can include or communicate with knowledge base 1414, which includes filter configuration information, filter kernel parameter selection criteria, filter kernel Parameters and / or other information useful for the deployment, configuration, and application of adaptive and steerable filters are stored. For example, knowledge base 1414 may store information that associates one or more filter kernel configurations, parameters, etc. with a particular structure that can be identified in the image signal. In addition or alternatively, knowledge base 1414 may store information used in identifying specific structures, structure orientations, and the like.

  Sophisticated knowledge bases in which information or portions thereof are indexed or otherwise accessible within the context can be utilized in accordance with embodiments of the present invention. For example, the diagnostic ultrasound system 1400 can be used for a plurality of predefined procedures or modes of operation such as a cardiac scan, a kidney scan, an upper gastrointestinal scan, and the like. The knowledge base 1414 can store information tailored to or specific to various of these procedures or modes of operation, allowing the user to use it in a selected one of the procedures. When configuring a diagnostic ultrasound system 1400 for that purpose, the relevant portion of the knowledge base 1414 may be adapted to the specific structure typical of such a procedure, the structural orientation typical of such a procedure, and such a procedure. Access is made to obtain information for identifying one or more tuned filter kernel configurations, filter parameters tuned for such a procedure, and the like. For example, a feature can be identified in an image signal, such as by using the fuzzy logic described above, and the knowledge base is accessed to the specific filter kernel and / or filter parameters used for that feature. Can be selected. If there is prior knowledge as to which features are likely to be present in the image signal (eg, via the selected mode of operation or the specific procedure being performed), the information can be used for feature identification and / or Or it can be taken into account in the filter selection decision. Of course, the knowledge base 1414 of the embodiments may additionally or alternatively be tailored to the particular context in other ways, such as addressing information with broader applicability or non-predetermined users. Information that may not be included.

  FIG. 15 shows a functional block diagram of a processing unit, such as that that may correspond to the external DSP 1404 of FIG. 14, in accordance with various embodiments of the invention. In the illustrated embodiment, the preprocessing component 1500 performs processing performed on the input image data, such as a prefiltering, mapping process, or other processes performed prior to the adaptive and steerable filtering of the present invention. provide.

  After preprocessing, the image signal is decomposed into a multi-resolution representation (sub-image) of the image by decomposition block 1501 in the illustrated embodiment. One embodiment can have up to N sub-images, so that the input image can be decomposed into N sub-images (the original image from which other sub-images are decomposed, It should be understood that it can be included as a “sub-image” for filtering as described herein). For example, decomposition block 1501 starts with a high resolution image signal, decomposes the signal into a first decomposed image signal that is half the resolution of the first signal, and converts the first decomposed image signal to the first Decompose into a second decomposed image signal at half the resolution of the decomposed image signal (1/4 of the resolution of the first signal), and so on, each having half the resolution of the next sub-image N sub-images can be provided. For example, consider an image with 128 pixels in each dimension, the next level of resolution should be 64x64, then 32x32, then 16x16, then 8x8, etc. . Decomposition according to preferred embodiments of the present invention is accomplished from top down (eg, from highest resolution to lowest resolution).

  It should be appreciated that the present invention is not limited by the number of levels of resolution or the level of decomposition between sub-images. Similarly, the present invention is not limited by the form of decomposition. Thus, various methods of decomposing an image can be used, including wavelet decomposition and various other forms now known or later developed.

  The concept of multi-resolution image processing according to the present invention can be demonstrated by the human eye. If the viewer is standing 10 feet away from the image, the resolution will be lower and what is seen is the structure or global features in the image. However, if the viewer is close, for example one foot away from the image, the viewer will see more details in the image, perhaps at the expense of looking at global features.

  Different levels of abstraction identify different features within the level of abstraction, such as global features, more localized features, and highly localized features, and apply filtering to them. Implemented in accordance with embodiments of the present invention. For example, embodiments can detect feature sides, and different feature sides apply speckle reduction filters on different feature sides. The system can use a lower resolution sub-image to extract global features of the image, and use a higher resolution sub-image to extract details to be stored.

Referring once again to FIG. 15, decomposition block 1501 provides decomposition of the image into sub-images for filtering. The output of decomposition block 1501 is illustrated as H ₀ to L _n−1 in the illustrated embodiment, but is provided to processing block 1502, which may perform adaptive filtering and / or according to embodiments of the invention. Provide steerable filtering. Accordingly, it is within processing block 1502 that the previously described filter is applied to the image in accordance with the illustrated embodiment.

  The processing block 1502 of the preferred embodiment demonstrates a kind of dependency. That is, in addition to the respective sub-images supplied to the processing block, information about features from lower resolution blocks, if available, is also supplied to the processing block (eg, processed by processing block 1502a). Information about the features is propagated to processing block 1502b). This additional information provides a basis from the lower resolution sub-image, which guides the processing performed on the higher resolution sub-image. Accordingly, preferred embodiments of the present invention provide image filtering using processing block 1502 from the bottom up (eg, from lowest resolution to highest resolution). Such bottom-up processing provides advantages and processing economics in identifying global features and in entering localized and highly localized features.

The processed sub-images output by processing block 1502 are illustrated as P ₀ through P _N−1 in the illustrated embodiment, but are provided to reconstruction block 1503 for multi-resolution image reconstruction. That is, the reconstruction block 1503 of the embodiments results in a combination of sub-images (eg, the opposite of the decomposition performed). The system of the illustrated embodiment intelligently combines the outputs from individual processing blocks 1502 to create a filtered image for a human user. Image reconstruction according to embodiments of the present invention can perform upsampling and combination. For example, a lower resolution sub-image can be upsampled to the resolution of the next higher resolution sub-image and the two sub-images can be combined. Such upsampling and combination can be repeated until the resolution of the original image is reached. Accordingly, preferred embodiments of the present invention provide image reconstruction using reconstruction block 1503 from the bottom up (eg, from lowest resolution to highest resolution). It should be understood, however, that the invention is not limited by the combination of shapes, as all currently known or later developed shapes can be used in one or more embodiments.

  Post processing component 1504 can be used to provide the desired additional signal processing after image reconstruction. For example, it may be desirable to increase brightness, remap greyscale, perform additional filtering to take care of medical processing, etc. after an image has been processed according to the present invention. Thus, the post processing provided by post processing component 1504 is typically a small component of the overall signal processing.

  As described above with respect to FIG. 14, the knowledge base 1414 can be used to store various application specific data. The processing performed in processing block 1502 often depends on features, various attributes expected on the image, etc., and therefore knowledge base 1414 provides useful information to tailor the processing to the various features. can do. It may be desirable to provide some control over this process, depending on the type of scanhead used, the type of sending application used, etc. For example, cardiology may use heart disease specific equipment. This process application specific adjustment is performed using information from the knowledge base 1414, which provides additional parameters utilized by the processing block. The knowledge base can include, for example, previous knowledge about the composition of available images. For example, if the image is known to be an image of the heart, embodiments can apply specialized algorithms that make the application better for cardiac imaging.

  FIG. 16 shows details regarding an embodiment of the decomposition block 1501. In the embodiment of FIG. 16, a decimation filter is used to create a sub-image having half the resolution of the next higher-order image or sub-image. For example, a decimation filter 1600 is used on the original noisy image signal to produce a sub-image 1602 having half resolution. This sub-image is used as an input to the next decimation filter that creates another sub-image. This sub-image is also used as an input to the interpolation filter 1601 (shown as a sub-image 1603). This interpolation filter is used to reconstruct a smooth version of the original image. This smoothed version is subtracted from the original image to produce a high pass version of the original image.

  FIG. 17 shows details regarding an embodiment of processing block 1502. As noted above, processing block 1502 is where the adaptive filter of embodiments of the present invention is applied. Therefore, the input signal is one of the sub-images. Local features are extracted by feature extraction block 1701 from this sub-image input and, if available, information about features from lower resolution sub-image processing. Information about the features is provided to the upsampler block 1704 by the feature extraction block 1701 of the illustrated embodiment to provide feature information to the processing blocks used for higher resolution sub-images. Information about the features is also supplied to the filter construction block 1703 by the feature extraction block 1701. The filter construction block 1703 preferably uses one or more adaptive filters and / or applied to the sub-image as detailed above using the feature information in combination with information available from the knowledge base. A steerable filter is selected, configured and / or calculated. One or more filters determined by the filter construction block 1703 are applied to the sub-image by the filtering block 1702. It should be appreciated that the adaptive filter and / or space-time filter applied by the filtering block 1702 need not be a single filter. For example, a symmetric filter followed by a cascade of asymmetric filters can be applied to the sub-image according to embodiments of the present invention.

  As can be seen, processing block 1502 of FIG. 17 outputs information (if appropriate) about the extracted features used in the next higher resolution processing block, and the next lower resolution processing. Information about the features from the block (if appropriate) is provided to processing block 1502. Information about features of lower resolution blocks is preferably used with information about features extracted at the current resolution level to calculate or otherwise select adaptive filter coefficients. In this example, the features used are those from two adjacent resolution levels, which can help achieve consistency between different processing blocks.

  FIG. 18 shows details regarding an embodiment of the reconstruction block 1503. The embodiment of FIG. 18 shows how the outputs of sub-image processing block 1502 are combined to form a composite image. Specifically, the remapping block 1801 allows image brightness to be readjusted as needed. The mapping function used here is obtained from the knowledge base 1414 of FIG. The output of the remapping block 1801 is provided to the upsampler 1802 for upsampling of the sub-image to the next higher sub-image resolution. A combiner 1803 then combines the upsampled sub-image with the next higher sub-image, and so on.

  FIG. 19 illustrates an exemplary DSP hardware configuration adapted in accordance with one or more embodiments to provide the functional blocks described above with respect to external DSP 1404. In the illustrated embodiment, an I / O for interfacing the DSP with other circuitry, such as for receiving a noisy image signal, supplying an output of a filtered image signal, and interfacing with a knowledge base 1414. An O port 1904 is included. The DSP 1404 in the illustrated embodiment includes a DMA engine 1903, a frame buffer 1905, a high speed memory 1902, and a DSP core 1901. The DSP 1404 in the illustrated embodiment includes a DMA engine 1903, a frame buffer 1905, a high speed memory 1902, and a DSP core 1901. The DSP core 1901 includes an arithmetic logic unit (ALU). High speed memory 1902 provides memory used by DSP core 1901 during computation. The DMA engine 1903 facilitates background data transfer between the high speed memory 1902 and an external, usually slower memory (with a frame buffer 1905). Although a particular configuration of a DSP is shown in FIG. 19, it should be appreciated that the present invention is not limited to a particular DSP or other type of processor to implement the multi-resolution adaptive filtering described above. Indeed, such processing can be performed, for example, by a field programmable gate array (FPGA), application specific integrated circuit (ASIC), general purpose microprocessor, or the like.

  An advantage of some embodiments is that the adaptive filter can provide better resolution by preserving edges. When combined with multi-resolution decomposition, it may be possible to perform high performance processing more efficiently. Another advantage of some embodiments is that multi-resolution processing can provide a more efficient way to extract information from high-level features to low-level details from the signal. Indeed, some embodiments can be implemented on portable devices because more efficient processing provides higher performance with lower power usage and less computing capabilities.

  Many of the calculations used for imaging solve differential equations. For example, if a dominant feature that spans a large area of an image must be extracted, using a traditional processing device may create a very large filter. However, various embodiments of the present invention break down the signal into lower resolution sub-images so that fewer pixels or points are processed so that features can be identified using smaller kernels. Become. For example, assume a 30 × 30 or 50 × 50 kernel adapted to different pixels. However, using the inventive concept where multi-resolution decomposition is used, smaller filter kernels can be used, for example 3x3 or 5x5. Additional performance enhancements can be achieved through the use of pre-calculated look-up tables and the use of simplified techniques, such as for processing block 1502. High performance filtering with lower power usage and lower costs can provide high quality portable imaging devices, such as ultrasound devices.

  Having described the invention and its advantages in detail, it will be understood that various changes, substitutions, and modifications may be made without departing from the spirit and scope of the invention as defined by the appended claims. I want. Furthermore, the scope of the present application is not intended to be limited to the specific embodiments of the processes, machines, products, material compositions, means, methods, and steps described herein. Those skilled in the art, from the disclosure of the present invention, will now have existing or later developed processes that perform substantially the same functions or achieve substantially the same results as the corresponding embodiments described herein. It will be readily appreciated that any machine, product, material composition, means, method, or step may be utilized in accordance with the present invention. Accordingly, the appended claims are intended to include within their scope such processes, machines, manufacture, compositions of materials, means, methods, or steps.

FIG. 6 illustrates application of an adaptive filter and / or steerable filter to an image according to an embodiment of the present invention. FIG. 3 is a diagram illustrating two-dimensional signal processing using intra-frame processing according to an embodiment of the present invention. FIG. 3 is a diagram illustrating two-dimensional signal processing using interframe processing according to an embodiment of the present invention. It is a figure which shows the three-dimensional signal processing by embodiment of this invention. FIG. 5 illustrates a noisy step function signal that can be filtered according to an embodiment of the present invention. 4B illustrates application of adaptive filtering to the noisy step function signal of FIG. 4A, according to an embodiment of the present invention. FIG. FIG. 4B is a diagram illustrating the application of a normal non-adaptive filter to the noisy step function signal of FIG. 4A. It is a figure which shows how the adaptive filter of embodiment of this invention can operate | move in one-dimensional space. FIG. 7 shows a two-dimensional case where adaptive filtering is applied similar to the one-dimensional case of FIG. FIG. 7 shows a two-dimensional case where adaptive filtering is applied similar to the one-dimensional case of FIG. FIG. 6 illustrates an exemplary symmetric adaptive filter applied to a ridge feature according to an embodiment of the present invention. FIG. 6 illustrates an exemplary symmetric adaptive filter applied to a ridge feature according to an embodiment of the present invention. FIG. 6 illustrates an exemplary symmetric adaptive filter applied at feature intersections according to embodiments of the invention. FIG. 6 illustrates an exemplary symmetric adaptive filter applied at feature intersections according to embodiments of the invention. FIG. 6 illustrates various example steerable filters used in at least one embodiment of the present invention. FIG. 6 illustrates various example steerable filters used in at least one embodiment of the present invention. It is a figure which shows the graph expression of the filter of embodiment of this invention. FIG. 4 illustrates edge and gradient directions in an image according to various embodiments of the invention. FIG. 4 illustrates a simple example embodiment of a Gaussian filter kernel that can be utilized in accordance with an embodiment of the present invention. FIG. 4 illustrates a simple example embodiment of a Gaussian filter kernel that can be utilized in accordance with an embodiment of the present invention. FIG. 4 illustrates a simple example embodiment of a Gaussian filter kernel that can be utilized in accordance with an embodiment of the present invention. FIG. 6 illustrates an exemplary signal path of a diagnostic ultrasound system adapted for use with various embodiments of the present invention. FIG. 6 is a functional block diagram illustrating processing units that can be used in image filtering according to various embodiments of the invention. FIG. 16 shows details regarding the embodiment of the exploded block of FIG. 15. FIG. 16 is a diagram illustrating details regarding the processing block embodiment of FIG. 15. FIG. 16 illustrates details regarding the embodiment of the reconfiguration block of FIG. 15. FIG. 2 illustrates an exemplary digital signal processor hardware configuration adapted according to embodiments of the present invention.

Explanation of symbols

100 images 111-114 circles 121-127 ellipses 201-209 frames 211-214 groups 310 and 302 blocks 401 step function signal 402 noise signal 403 noisy step function signal 404 filtered step function signal 410 filter 504 filtered step Function signal 510 Filter kernel 521 and 522 Regions 611-615 Filter kernel 703 Noisy step function signal 704 Filtered step function signal 711 Filter kernel 712 Filter kernel 803 Noisy ridge signal 804 Filtered ridge signal 811 Filter kernel 812 Filter kernel 903 Noisy ridge crossing signal 904 Filter Ringed ridge crossing signal 911 Filter kernel 1400 Diagnostic ultrasound system 1401 Scan head 1402 Front end network 1403 Signal processor 1404 External DSP
1405 Back-end network 1406 Display 1414 Knowledge base 1500 Preprocessing component 1501 Decomposition block 1502 Processing block 1503 Reconstruction block 1504 Postprocessing component 1600 Decimation filter 1601 Interpolation filter 1602 Sub image 1603 Sub image 1701 Feature extraction block 1702 Filtering Block 1703 Filter configuration block 1704 Upsampler block 1801 Remapping block 1802 Upsampler 1803 Combiner 1901 DSP core 1902 High speed memory 1903 DMA engine 1904 I / O port 1905 Frame buffer

Claims (47)

A method of processing an image comprising:
Decomposing the image into a plurality of sub-images;
Determining one or more features in each sub-image of the plurality of sub-images;
Applying an adaptive filter group separately to each sub-image of the plurality of sub-images, wherein the adaptive filter group is adaptive to the one or more related feature aspects; Including methods.   The method of claim 1, wherein the plurality of sub-images include sub-images of different resolutions.   The method of claim 2, wherein the sub-images of the plurality of sub-images are each half the resolution of the next higher-resolution sub-image.   The method of claim 2, wherein the decomposition is achieved from a highest resolution to a lowest resolution of the different resolutions. The determination of one or more features in each sub-image comprises:
The method of claim 1, comprising accepting feature information about another sub-image of the plurality of sub-images for use in determining one or more features of a particular sub-image.   6. The method of claim 5, wherein the determination of one or more features in each sub-image is accomplished from a lowest resolution to a highest resolution. The determination of one or more features in each sub-image comprises:
The method of claim 1, comprising identifying feature edges in the sub-image. The determination of one or more features in each sub-image comprises:
The method of claim 1, comprising identifying gradients in the sub-image. The determination of one or more features in each sub-image comprises:
The method of claim 1, comprising accessing a knowledge base that stores information about image features associated with a particular mode of operation of the host system. The determination of one or more features in each sub-image comprises:
The method of claim 1, comprising accessing a knowledge base that stores information about image features associated with a particular procedure performed using a host system. The application of the adaptive filter group is
The method of claim 1, comprising using interframe information in filtering sub-image features. The application of the adaptive filter group is
The method of claim 1, comprising using intraframe information in filtering sub-image features.   The method of claim 1, wherein the adaptive filter group is adaptive with respect to a spatial aspect of associated features of the sub-image.   The method of claim 1, wherein the adaptive filter group is adaptive with respect to temporal aspects of associated features of the sub-image.   The method of claim 1, wherein the adaptive filter group is adaptive with respect to edges of associated features of the sub-image.   The method of claim 1, wherein the adaptive filter group is adaptive with respect to a slope of an associated feature of the sub-image.   The method of claim 1, wherein the adaptive filter group is adaptive with respect to a gradient of associated features of the sub-image. The method of claim 1, further comprising: determining one or more filter parameters used for the application of adaptive filters as a function of the one or more features. The determination of the one or more filter parameters comprises:
19. The method of claim 18, comprising accessing a knowledge base that stores information about the features. The method of claim 1, further comprising determining an orientation of at least one of the one or more features. The application of the adaptive filter group is
21. The method of claim 20, comprising steering an adaptive filter of the adaptive filter group as a function of the orientation of the at least one feature.   21. The method of claim 20, wherein the orientation includes a spatial orientation.   21. The method of claim 20, wherein the orientation includes a temporal orientation. The method of claim 1, further comprising: reconstructing a filtered image from the plurality of sub-images after applying adaptive filter groups to each of the sub-images separately.   25. The method of claim 24, wherein the decomposition of the image, the application of adaptive filter groups, and the reconstruction of the filtered image are performed multiple times.   26. The method of claim 25, wherein the multiple times are associated with introducing a new processing point for the image. A method of processing an image comprising:
Decomposing the image into a plurality of sub-images;
Determining one or more features in each sub-image of the plurality of sub-images;
Determining one or more adaptive filter parameters as a function of the one or more features;
Separately applying an adaptive filter group to each sub-image of the plurality of sub-images, wherein the adaptive filter group is adaptive to the one or more related feature aspects, the adaptive filter group Performing and applying one or more of the adaptive filter parameters and re-filtering the filtered images from the plurality of sub-images after applying adaptive filter groups to each of the sub-images separately. A method comprising: configuring.   28. The method of claim 27, wherein the sub-images include sub-images of different resolutions, each different resolution being half of the next higher resolution.   28. The method of claim 27, wherein the decomposition of the image is accomplished from a highest resolution to a lowest resolution and the reconstruction of the filtered image is accomplished from a lowest resolution to a highest resolution. The determination of one or more features in each of the sub-images;
Determining the first one or more features in the lowest resolution sub-image;
Providing information about the one or more features to a process for determining one or more features in a higher resolution sub-image;
28. using the information about the first one or more features to determine a second one or more features in the higher resolution sub-image. Method. The determination of the one or more adaptive filter parameters comprises:
28. The method of claim 27, comprising accessing a knowledge base that stores information about the features.   28. The method of claim 27, wherein the adaptive filter group is adaptive with respect to a spatial aspect of associated features of the sub-image.   28. The method of claim 27, wherein the adaptive filter group is adaptive with respect to temporal aspects of associated features of the sub-image. 28. The method of claim 27, further comprising: determining an orientation of at least one of the one or more features. The application of the adaptive filter group is
35. The method of claim 34, comprising steering an adaptive filter of the adaptive filter group as a function of the orientation of the at least one feature. A system for processing images,
A multi-resolution image decomposer operable to receive image data and create a plurality of sub-images therefrom, each of the sub-images having a different resolution;
A plurality of each operable to determine one or more features in related sub-images of the plurality of sub-images and provide filtering of the related sub-images as a function of the one or more features A plurality of processing blocks, wherein one or more of the processing blocks receives information regarding characteristics determined for another sub-image by another processing block of the plurality of processing blocks;
An image reconstructor operable to receive output from the plurality of processing blocks and to produce a combined image therefrom.   40. The system of claim 36, wherein the plurality of processing blocks are provided by a digital signal processor.   38. The system of claim 37, wherein the multi-resolution image decomposer and the image reconstructor are provided by the digital signal processor. The system of claim 36, further comprising a database that stores filter parameter information for use by the plurality of processing blocks.   40. The system of claim 39, wherein the filter parameter information is associated with a particular feature as determinable by the processing block.   40. The system of claim 39, wherein the filter parameter information relates to a specific mode of operation of the system.   40. The system of claim 39, wherein the filter parameter information is associated with a particular procedure performed using the system.   40. The system of claim 36, wherein the filtering comprises adaptive filtering.   44. The system of claim 43, wherein the adaptive filtering includes adaptation of one or more filter parameters as a function of a spatial aspect of an associated feature of the features.   44. The system of claim 43, wherein the adaptive filtering includes adaptation of one or more filter parameters as a function of temporal aspects of related features of the features.   40. The system of claim 36, wherein the filtering comprises steered filtering.   48. The system of claim 46, wherein the steered filtering includes steering the filter as a function of the orientation of the related features of the features.

Images