KR100740646B1 - Bit rate control method in still image compression method and apparatus therefor - Google Patents

Abstract

The present invention relates to a method for controlling a bit rate in a still image compression method and an apparatus therefor, the method comprising: a discrete cosine transform step (a) of calculating a discrete cosine transform coefficient for still image data consisting of horizontal and vertical blocks (a)-the discrete cosine Transform (DCT) coefficients include DC coefficients and AC coefficients; (B) quantizing the calculated discrete cosine transform coefficients; (C) zigzag scanning the quantized coefficients, wherein zigzag scanning means one-dimensional ordering from low frequency components to high frequency components sequentially; (D) an entropy encoding step of allocating a code of a predetermined bit to the one-dimensionally arranged quantization coefficients; (E) checking a remaining capacity of an output buffer for temporarily storing the code; And (f) removing the code corresponding to the AC coefficient in the reverse order of the zigzag scanning when the remaining capacity of the output buffer exceeds a preset value. will be. According to the present invention, since the bit rate of the still image is controlled according to the zigzag scanning characteristic, an efficient bit rate control is possible.

Bit Rate, Jpeg, Zigzag Scanning, MCU, Encoding

Description

TECHNICAL FIELD OF THE INVENTION A BIT RATE CONTROL METHOD IN A STILL IMAGE COMPRESSION METHOD AND A DEVICE FOR THE SAME

1 is a block diagram showing the configuration of a still image compression apparatus according to the prior art.

2 is a diagram illustrating a DC coefficient in a DCT coefficient block calculated according to a general DCT process.

3 is a diagram illustrating zigzag scanning in a general still image compression process.

4 is a diagram illustrating a stream format system in a general still image compression method.

5 is a diagram showing the configuration of a still image compression apparatus according to an embodiment of the present invention.

6A to 6C are diagrams illustrating a bit rate control process according to an exemplary embodiment of the present invention.

7 illustrates a minimum code unit in which bit rate control is performed according to the present invention;

The present invention relates to a method for controlling a bit rate in a still image compression method and an apparatus therefor, and more particularly, to a bit rate control in a still image compression method capable of efficiently adjusting the amount of transmission data of a still image compressed in a JPEG compression A method and apparatus therefor.

Recently, with the high integration of image sensors such as Charged Coupled Device (CCD) sensors and Complementary Metal Oxide (CMOS) sensors and the development of digital image processing technology, digital cameras are rapidly evolving, and the use of digital cameras is rapidly increasing. to be.

The CCD sensor and the CMOS sensor include a plurality of pixels, and convert the light signal of the subject into RGB data when the image is captured, and the RGB data is composed of luminance (Y) and saturation (Cb and Cr) information. Is converted into YUV data (or YIQ data).

When the data thus converted is stored in a memory or transmitted to another user through a network, a compression process of image data is performed to reduce the data amount.

1 is a block diagram showing the configuration of an image compression apparatus according to the prior art.

A compression of a still image according to the prior art will be described in detail with reference to FIG. 1.

As shown in FIG. 1, the image compression apparatus according to the related art includes a sampling unit 100, a DCT unit 102, a quantizer 104, and an entropy encoder 106.

In the Joint Photographic Experts Group (Jpeg) file, YUV data, Y, Cb, and Cr, which are components of YUV data, may be stored in various ratios. The Y component of every pixel can be stored while Cb and Cr can be skipped by one pixel.

As described above, the sampling unit 100 performs a process of selecting and storing only data of some pixels from all pixels, and the sampled data is output to the DCT unit 102.

At this time, the sampled data is divided into 8x8 blocks, which is called a MCU (Minimum Coded Unit).

As described above, data divided into 8x8 blocks is input to the DCT unit 102, and the DCT unit 102 performs a discrete cosine transform process for converting image data from the time domain to the frequency domain.

Through the discrete cosine transform process, sample data of an 8x8 block is transformed into DCT coefficients of an 8x8 block, where the DCT coefficients represent DC coefficients representing energy of the low frequency spectrum and the remaining high frequency spectrum, as shown in FIGS. It includes the AC coefficient that it represents.

Meanwhile, the DCT unit 102 outputs the calculated DCT coefficients of the 8 × 8 block to the quantizer 104, and the quantizer 104 performs a process of quantizing through a pre-stored quantizer table.

The quantization process is a process of dividing a DC coefficient and an AC coefficient by a value stored in a quantization table, and then rounding them. Through this quantization process, most AC coefficients are inherently converted to zero.

Thereafter, as illustrated in FIG. 3, the quantized coefficients are zigzag scanned and output to the entropy encoder 106, and the coefficients transformed in one dimension by zigzag scanning undergo entropy coding.

In general, entropy coding uses Run Length Coding (RLC) or Variable Length Coding (VLC). The entropy coding is performed by sequentially applying a code (code word) to a coefficient input after being zigzag scanned using the above method. ) Process.

Through this, as shown in FIG. 4, the finally compressed data 108 for each MCU is output, and the compressed data is transmitted to the outside through the network after filling the output buffer.

In MPEG, which is a video compression method, it is common to control the bit rate by applying various methods such as changing the quantization coefficients in front of the macro block stream to control the bit rate. It is common that such a bit rate control process is not performed, thereby causing a change in image quality of a transmitted image.

On the other hand, there is a case of using a method of changing the quantization table applied by the quantizer 104 in order to control the amount of data in the still image compression method in the prior art, a conventional quantizer generally uses about two quantization tables Because of this, it cannot cope with various conditions, and thus there is a problem that efficient bit rate control cannot be performed.

In order to solve the problems of the prior art as described above, the present invention proposes a method for controlling a bit rate in a still image compression method and an apparatus therefor that can perform efficient bit rate control when compressing still images.

In order to achieve the above object, according to a preferred embodiment of the present invention, the discrete cosine transform step (a)-calculating the discrete cosine transform coefficients for still image data consisting of horizontal and vertical blocks Transform (DCT) coefficients include DC coefficients and AC coefficients; (B) quantizing the calculated discrete cosine transform coefficients; (C) zigzag scanning the quantized coefficients, wherein zigzag scanning means one-dimensional ordering from low frequency components to high frequency components sequentially; (D) an entropy encoding step of allocating a code of a predetermined bit to the one-dimensionally arranged quantization coefficients; (E) checking a remaining capacity of an output buffer for temporarily storing the code; And (f) removing a code corresponding to the AC coefficient in a reverse order of the zigzag scanning when the remaining capacity of the output buffer exceeds a preset value. do.

Here, the step (f) is preferably performed in the MCU unit of 8x8 block, the MCU unit is preferably identified by the End Of Block (EOB) marker.

The method may further include maintaining one or more levels in which the number of removal of codes corresponding to the AC coefficients is set in advance. The step (f) may include codes corresponding to the AC coefficients by a number set according to the one or more levels. Can be removed.

On the other hand, according to another aspect of the present invention, a discrete cosine transform means for calculating a discrete cosine transform coefficient for a still image, wherein the discrete cosine transform coefficient includes a DC coefficient and an AC coefficient; Quantization means for quantizing the calculated discrete cosine transform coefficients; Scanning means for zigzag scanning the quantized coefficients, said zigzag scanning means one-dimensional arrangement sequentially from a low frequency component to a high frequency component; Entropy encoding means for allocating a code of a predetermined bit to the one-dimensionally arranged quantization coefficients; And a bit rate control means for checking the remaining capacity of the output buffer for temporarily storing the code, and if the remaining capacity of the output buffer exceeds a preset value, removing the code corresponding to the AC coefficient in the reverse order of the zigzag scanning. Provided is a bit rate control apparatus in a still image compression method including a.

Hereinafter, with reference to the accompanying drawings will be described in detail a preferred embodiment of the bit rate control method and apparatus therefor in the still image compression method according to the present invention.

The present invention is to control the bit rate of the transmission of the compressed still image through the network, not only when transmitting a still image via a wireless network in a mobile, such as a digital camera, but also when transmitting a still image through a wired network Applicable to

According to a preferred embodiment of the present invention, the bit rate control may be performed by an application program for still image compression, or alternatively, a hardware configuration having a predetermined circuit structure may be achieved.

In addition, the bit rate control method proposed in the present invention may be implemented in various forms, and it will be apparent to those skilled in the art that simple modifications of the embodiments fall within the scope of the present invention.

5 is a diagram illustrating a configuration of an image compression apparatus according to an exemplary embodiment of the present invention.

As shown in FIG. 5, the image compression apparatus according to the present invention may include a sampling unit 500, a DCT unit 502, a quantizer 504, an entropy encoder 506, and a bit rate controller 508. .

As described above, the RGB data captured by the image sensor is converted into YUV data and input to an image compression device (codec: codec), and the sampling unit 500 preconstructs the Y, Cb, and Cr components of the YUV data in advance. Sampling is performed for each set sampling period.

In the present specification, a process of controlling a bit rate based on YUV data will be described, but it is not necessarily limited thereto, and it should be understood that the case where RGB data is converted into YIQ data is also included in the scope of the present invention.

The sampled data should be divided into a predetermined block size for the DCT process. As described above, this is called an MCU, and the MCU generally means an 8x8 block.

Here, the pixel block size of the MCU may vary according to the sampling period of the sampling unit. For example, when Y: Cb: Cr is 1: 1: 1, the MCU selects the components included in each pixel, so that the MCU It will be 8x8 pixels, with a 1: 1: 1, i.e. Y for every pixel, and Cb and Cr will be skipped one by one, and the MCU will be 16x16 pixels.

The sampled data is input to the DCT unit 502, and the DCT unit 502 converts a sample of an 8x8 block into DCT coefficients.

The data input to the DCT unit 502 includes amplitude information of each pixel included in the MCU, wherein the discrete cosine transform process separates amplitude changes of adjacent pixels into eight frequency spectrums and converts energy of each frequency spectrum. This is the process of calculating the DCT coefficients quantitatively displayed.

In the above, since the MCU is an 8x8 block, 64 DCT coefficients are calculated for one MCU, and such DCT coefficients include one DC coefficient and 63 AC coefficients for each MCU.

The DCT unit 502 arranges the calculated DCT coefficients in the form of 8x8 blocks according to the MCU. As shown in FIGS. 6A to 6C, the DC coefficients are located in the upper left corner, and the DCT coefficients in the lower frequency component toward the upper left corner. Will be located.

The DC coefficient represents the lowest frequency component and has a value relatively larger than the AC coefficient.

The DC coefficient is larger than the AC coefficient because most of the parts representing objects appearing at low frequencies are included in the overall view of the image, and a few of the contour parts appearing at high frequencies are included.

delete

The DCT coefficients of the 8x8 block calculated by the DCT unit 502 are input to the quantizer 504 and converted into quantization coefficients.

Quantization refers to the process of rounding DCT coefficients to integers within ranges that are indistinguishable from human senses. In this process, the DCT coefficients obtained through the DCT transform are divided into predefined constants (constants in the quantization table), and the result is rounded to an integer value. The constants used here are greater than or equal to 1 and all of them vary in frequency and height.

In other words, the low frequency to which the human eye responds sensitively should be divided into small constants so that there is no big difference from the original value. do.

After the actual quantization process, the low frequency coefficients are non-zero and the high frequency coefficients are almost zero.

The quantization process is the portion where the largest data compression occurs and the most data loss at the same time. Quantization is performed through one or more quantization tables, but only a small number of quantization tables are used depending on the data compression rate. .

The quantization coefficients are output to the entropy encoder 506 through zigzag scanning. Although the zigzag scanning means is not illustrated in FIG. 5, such means may be included in the quantizer 504.

The entropy encoder 506 performs final compression using lossless compression as the last step of the JPEG algorithm. In general, the entropy encoder 506 uses a lot of Huffman coding.

Huffman coding assigns short-length codes (short bits) to high-frequency values (quantization coefficients) and long-length (long bits) codes to low-frequency values, reducing the overall length. That's the way.

Many of the coefficients become zero through quantization, and the same shape can appear as many as 64 digits through zigzag scanning of these coefficients. Therefore, they are compressed by Huffman coding to reduce their size.

As such, a table for the allocated Huffman code is stored together in the compressed file, and when the data is decoded, a process of converting the original Huffman code into the original data may be performed.

Conventionally, as described above, the entropy coded data is filled in an output buffer (not shown) without a process for separate bit rate control and is transmitted to the outside through a transmission end (not shown) connected to a network. The still image compression apparatus may further include a bit rate controller 508.

The bit rate controller 508 performs a process of additionally removing a code corresponding to a high frequency component among the codes (code words) allocated by the entropy encoder 506. The bit rate controller 508 checks the amount of data filled in the output buffer and performs entropy coding. If it is determined that the amount of data stream is to be reduced, the process of reducing the amount of stream is performed by using the zigzag scanning characteristic of the jpeg compression.

As described above, the DC coefficient is located at the upper left of the DCT coefficient of the 8x8 block obtained through the DCT process, and the AC coefficient for the higher frequency component is located toward the lower right (see FIGS. 3 and 6A). The DCT coefficients included in the 8x8 block by zigzag scanning are sequentially one-dimensionally arranged from the low frequency components to the high frequency components.

As described above, since the high frequency components are sequentially output from the low frequency components by the zigzag scanning in the JPEG compression process, when the bit rate control unit 508 according to the present invention wants to reduce the amount of data streams, the single bit rate for each MCU is determined. The code corresponding to the AC coefficient outputted later among the codes corresponding to the coefficients is removed.

6A to 6C are diagrams illustrating a bit rate control process according to an exemplary embodiment of the present invention, and FIG. 6A shows an 8x8 coefficient block before bit rate control, and FIGS. 6B to 6C show a high level. The process of bit rate control is shown.

6A to 6B, the bit rate controller 508 according to the present invention checks the amount of the output buffer and does not have an integer value of 0 in the quantization process and is assigned to each quantization coefficient by the entropy encoder 506. Removes the code in the reverse order of zigzag scanning described above (see reference numerals 600 and 602), and if it is necessary to further reduce the amount of data, additionally removes the code corresponding to the AC coefficients in the zigzag scanning reverse order. The process will be carried out.

FIG. 7 is a diagram illustrating a minimum code unit (MCU) in which bit rate control is performed according to the present invention. FIG. 7 is a diagram illustrating quantization coefficients of an 8 × 8 block according to FIG. 6 in one dimension. Before losing, it indicates a process in which additional data is reduced by bit rate control toward the upper level.

That is, according to the present invention, the bit rate controller 508 determines the level to be applied according to the remaining capacity of the checked output buffer while maintaining one or more levels in which the number of removal of codes corresponding to the AC coefficients is preset. The process of removing the code corresponding to the AC coefficient by the number set according to the level is performed.

In FIG. 7, data0 through data9 mean one or more codes assigned to coefficients, and an end of block (EOB) means an end of an 8x8 block (MCU).

Since the entropy encoder 506 outputs codes in the order of data1 to data9 according to the zigzag scanning, the data output later corresponds to the code for the high frequency component, and the bit rate controller 508 according to the present invention is an MCU unit divided by EOB. The high frequency component is removed.

Meanwhile, the amount of the high frequency component removed by the bit rate controller 508 may be determined according to the remaining capacity of the output buffer, and it may be possible to remove the high frequency component within a range not exceeding the bit rate currently allowable by the output buffer. .

In the above, the case of Huffman coding in entropy coding has been described as an example, but it is not necessarily limited thereto. It will be apparent to those skilled in the art that the present invention can be applied to other coding schemes applied to JP.

Preferred embodiments of the present invention described above are disclosed for purposes of illustration, and those skilled in the art will be able to make various modifications, changes, and additions within the spirit and scope of the present invention. Additions should be considered to be within the scope of the following claims.

As described above, according to the present invention, since the bit rate of the entropy coded data is removed by using the zigzag characteristic in the JPEG compression method, the amount of the data stream can be efficiently reduced.

Claims (8)

Calculating a discrete cosine transform coefficient for still image data consisting of horizontal and vertical blocks (a), wherein the discrete cosine transform (DCT) coefficients include DC coefficients and AC coefficients; (B) quantizing the calculated discrete cosine transform coefficients; (C) zigzag scanning the quantized coefficients, wherein zigzag scanning means one-dimensional ordering from low frequency components to high frequency components sequentially; (D) an entropy encoding step of allocating a code of a predetermined bit to the one-dimensionally arranged quantization coefficients; (E) checking a remaining capacity of an output buffer for temporarily storing the code; And And removing the code corresponding to the AC coefficient in the reverse order of the zigzag scanning when the remaining capacity of the output buffer exceeds a preset value. The method of claim 1, The step (f) is a bit rate control method in a still image compression method is performed in MCU units of 8x8 block. The method of claim 2, The MCU unit is a bit rate control method in a still image compression method identified by the End Of Block (EOB) marker. The method of claim 1, Maintaining at least one level in which the number of removal of codes corresponding to the AC coefficient is preset; The step (f) is a method of controlling the bit rate in the still picture compression method to remove the code corresponding to the AC coefficient by a number set according to the one or more levels. Discrete cosine transform means for calculating a discrete cosine transform coefficient for the still image, the discrete cosine transform coefficient including a DC coefficient and an AC coefficient; Quantization means for quantizing the calculated discrete cosine transform coefficients; Scanning means for zigzag scanning the quantized coefficients, said zigzag scanning means one-dimensional arrangement sequentially from a low frequency component to a high frequency component; Entropy encoding means for allocating a code of a predetermined bit to the one-dimensionally arranged quantization coefficients; And A bit rate control means for checking the remaining capacity of the output buffer for temporarily storing the code, and removing the code corresponding to the AC coefficient in the reverse order of the zigzag scanning if the remaining capacity of the output buffer exceeds a preset value; Bit rate control apparatus in a still image compression method comprising. The method of claim 5, And the bit rate control means removes the code in units of 8x8 blocks of the MCU. The method of claim 6, The MCU unit is a bit rate control apparatus in a still image compression method identified by the End Of Block (EOB) marker. The method of claim 5, The bit rate control means stores information about one or more levels at which the number of removal of codes corresponding to the AC coefficients is set in advance, and determines the level to be applied according to the capacity of the checked output buffer and sets the number according to the levels. And a bit rate control apparatus in a still picture compression scheme that removes a code corresponding to the AC coefficient by an amount.

Images