KR101089725B1 - Method of designing threshold filter for lossless image compression, apparatus and method for lossless image compression using the filter - Google Patents

Abstract

PURPOSE: A method for designing a threshold filter for lossless image compression, and an apparatus and method for compressing an image without loss using the filter are provided to improve a compression ratio of image with large data. CONSTITUTION: A pre-processing unit(10) processes input image through a DC-level shifting and tiling. A conversion unit(20) converts the processed image to a frequency band. A threshold filtering unit(30) selectively outputs a pixel value by comparing a pixel value included in a divided frequency sub band with the coefficient of the threshold filter. An entropy encoding unit(50) generates a compressed image through an entropy encoding of the filtered image.

Description

Method for designing threshold filter for lossless image compression, apparatus and method for lossless image compression using the filter

The present invention relates to a method for designing a critical filter for lossless image compression, a lossless image compression apparatus and method using the filter, and more particularly, to a method for designing a critical filter for visual losslessness in band division coding of JPEG2000 and the same. The present invention relates to a lossless image compression device used, and an image compression device for diagnostic losslessness and a method thereof.

Recently, JPEG2000, a band division coding method using a discrete wavelet transform (DWT), has been adopted as an international standard. In the band division coding method, JPEG2000 is characterized by subband image signals belonging to a high frequency subband excluding a low frequency subband and statistical characteristics of a Laplace probability density function (L-pdf). Has JPEG2000 can support lossy and lossless compression at the same time. In case of lossy compression, the background and interest areas are separated and encoded by ROI (Region of Interest) mask. It has strengths in that it minimizes image distortion on the edge of the important part. Nevertheless, for JPEG2000 (meaning lossy compression), which has a high compression factor of 9-7 taps, the quantizer does not take into account visual distortion due to quantization loss, but only minimum Mean-Squared. Because it is designed to have an error, a sharp image distortion occurs in the background and the edge part in high compression.

The technical problem to be solved by the present invention is to solve the problem of sharp image distortion in the background and edges due to visual distortion due to quantization loss in the case of image compression having a high compression rate, relatively large data capacity in the case of diagnostic medical image Due to this, there is a lot of load on storage and transmission, and the method of designing a critical filter that can overcome the limitation that the reading of the diagnostic specialist is disturbed by the image quality distortion when compressing the image to solve it, and the lossless using the filter An image compression apparatus and method are provided.

In order to solve the above technical problem, a method of designing a critical filter for lossless video compression according to the present invention comprises: splitting an input image of N (N is a positive integer) frame into a plurality of bands and Calculating a standard deviation; Calculating an average and standard deviation of N frames from the calculated average and standard deviation; Determining a threshold parameter for each divided band by referring to a preset threshold table in view of visual loss of the input image due to compression; Calculating coefficients of a threshold filter using the calculated average and standard deviation of the N frames and the determined threshold parameter; And selectively updating the threshold table by comparing a band error according to the calculated coefficient of the threshold filter with an error according to reverse band division after threshold filtering based on pixel values of the divided band.

In order to solve the above other technical problem, a lossless image compression method according to the present invention, pre-processing the input image by performing DC-level shifting (tiling) and tiling (tiling); Band dividing the preprocessed image into a frequency band; Performing threshold filtering by selectively outputting the pixel value by comparing a pixel value belonging to the divided frequency sub-band with a coefficient of a threshold filter; And entropy encoding the critical filtered image according to a probability of a symbol in the image to generate a compressed image, wherein the coefficient of the threshold filter is input of an N (N is a positive integer) frame. Split the image into a plurality of bands, calculate the mean and standard deviation for each band, calculate the average and standard deviation of N frames from the calculated mean and standard deviation, and visually reduce the visual loss of the input image due to compression. The threshold parameter is determined for each of the divided bands by referring to a preset threshold table in consideration, and is calculated by performing an operation using the average and standard deviation of the calculated N frames and the determined threshold parameter.

In order to solve the another technical problem, the lossless image compression method according to the present invention, the threshold filtering using the quantization step size (step size) calculated from the statistical information and the band weight of the frequency band signal of the input image The method further includes quantizing the image.

On the other hand, the following provides a computer-readable recording medium recording a program for executing the above-described method for designing a critical filter for lossless image compression and the lossless image compression method using the filter on a computer.

In order to solve the other technical problem, the lossless image compression apparatus according to the present invention, a pre-processing unit for pre-processing the input image by performing DC-level shifting and tiling; A converting unit converting the preprocessed image into a frequency band and band-split it; A threshold filtering unit configured to compare the pixel values belonging to the divided frequency subbands with coefficients of a threshold filter and selectively output the pixel values by performing critical filtering; And an entropy encoder configured to generate a compressed image by entropy encoding the critical filtered image according to a probability of a symbol in the image, wherein the coefficient of the threshold filter includes a plurality of input images of N (N is a positive integer) frames. Split into three bands, calculate an average and standard deviation for each band, calculate an average and standard deviation of N frames from the calculated average and standard deviation, and consider the visual loss of the input image due to compression in advance. A threshold parameter is determined for each of the divided bands by referring to the set threshold table, and is calculated by performing an operation using the average and standard deviation of the calculated N frames and the determined threshold parameter.

In order to solve the another technical problem, the lossless image compression apparatus according to the present invention, quantizing the threshold filtered image using the quantization step size calculated from the statistical information and the band weight of the frequency band signal of the input image It further includes a quantization unit.

The present invention guarantees visual losslessness through a threshold filter calculated using statistical information on a subband subdivided from an input image and a threshold table, and quantizes an image using a quantization step size based on band weights. It is possible to improve the compression ratio for images with large data volumes, such as diagnostic medical images.

1 is a diagram illustrating statistical characteristics of a diagnostic medical image and a general image.
2 is a block diagram illustrating an image compression device that guarantees visual losslessness according to an embodiment of the present invention.
3 is a diagram illustrating a threshold filtering method adopted by embodiments of the present invention.
4 is a block diagram illustrating a method of training a critical filter in a lossless image compression apparatus according to an embodiment of the present invention.
5 is a flowchart illustrating a method of designing a critical filter in the lossless video compression apparatus of FIG. 4, according to an exemplary embodiment.
6 is a flowchart illustrating a series of performing operations for optimizing the threshold parameters of FIGS. 4 and 5.
FIG. 7 is a block diagram illustrating an image compression device that guarantees a diagnostic lossless method according to another embodiment of the present invention.
8 is a block diagram illustrating a quantization process in more detail in the image compression device of FIG. 7 according to another embodiment of the present invention.
9 is a diagram illustrating a quantization method in an image compression device that guarantees a diagnostic lossless method according to another embodiment of the present invention.

Before describing the embodiments of the present invention, the environment in which the embodiments are implemented will be briefly introduced, and related problem situations will be presented.

In general, diagnostic medical images have a high resolution and have a data capacity of 10 to 16 bits per pixel, which not only takes up a lot of space in storage, but also places a heavy load on the system for transmission. Therefore, in order to solve this problem of storage and transmission, the diagnostic medical image must be compressed within a range that does not affect the reading of the diagnostic specialist. In addition, when the diagnostic medical image is compressed using a lossy compression method including a quantizer, which is a conventional image compression method, a medical condition requiring accurate diagnosis may be lost or deteriorated due to image quality distortion caused by the quantizer. It is impossible to judge.

1 is a diagram illustrating statistical characteristics of a diagnostic medical image and a general image, in which image [A] illustrates an original image, and images [B], images [C], and images [D] are 11, 12, It is a figure which illustrates the statistical characteristic of the 13th band.

As described above, since the diagnostic medical image has statistical characteristics different from that of the general image, the visual distortion caused by the compression may also be more serious than the general image. When comparing the statistics of the medical image illustrated in FIG. 1 and the statistics of the LENA image, the diagnostic medical image has more values closer to '0' and '255' than the general image, while other values are widely distributed. I can see. In addition, it can be seen that the diagnostic medical image has many values close to '0' even in the high frequency component after band division. This is because the diagnostic medical CT image includes high frequency noises generated by systemic factors (such as electric field changes) and environmental factors (lighting) of the medical diagnostic equipment as compared with general images. Therefore, in view of the above problem situation, embodiments of the present invention to be described below are to propose a specialized diagnostic medical image compression technique that is different from the conventional image compression technique.

Hereinafter, embodiments of the present invention will be described in more detail with reference to the accompanying drawings. Like reference numerals in the drawings refer to like elements.

FIG. 2 is a block diagram illustrating an image compression apparatus that guarantees visual losslessness according to an embodiment of the present invention. The preprocessor 10, the converter 20, the threshold filter 30, and the entropy encoder 50 are illustrated in FIG. ). Among these, a detailed description of the configuration of the other components except the threshold filtering unit 30 blurs the essence of the present invention and briefly introduces only an overview thereof.

The preprocessor 10 pre-processes the input image by performing DC-level shifting and tiling.

The conversion unit 20 divides the preprocessed image into a frequency band and divides the band. Those skilled in the art can recognize that the transform unit 20 can be implemented through a method such as a discrete wavelet transform (DWT).

The threshold filtering unit 30 performs threshold filtering by selectively outputting pixel values by comparing the pixel values belonging to the frequency sub-band divided by the converter 20 with the coefficients of the threshold filter. The threshold filtering method adopted by the embodiments of the present invention will be described with reference to FIG. 3.

The threshold filtering unit 30 is a component introduced for visual losslessness, and determines the degree of loss occurring in the compression process with reference to the threshold table 33. That is, as shown in FIG. 3, the value of the signal on the x-axis (meaning a pixel value belonging to the divided frequency subband) is compared with a threshold value t _i (meaning the coefficient of the critical filter), and the comparison result. If the pixel value is smaller than the coefficient of the threshold filter, 0 may be output. Otherwise, the pixel value may be output as it is.

The definition of the critical filtering coefficient and the symbols and abbreviations to be used below are introduced as in Equation 1 below.

는 다음의 수학식 2와 같은 기능을 수행한다.If the pixel value and the optimal threshold coefficient belonging to the subband of the diagnostic medical image are g (x, y) and thr _i , respectively, the threshold filtering output

Performs the same function as Equation 2 below.

Referring to FIG. 3 described above, it can be seen that the output signal is divided into 0 and g (x, y) based on the threshold set by Equation 2.

This threshold filtering is performed through an optimal threshold coefficient with visual losslessness calculated using the statistical characteristics of the frequency band signals of the diagnostic medical laboratory images (in the embodiments of the present invention, mean mean and standard deviation). Meanwhile, a more specific method of calculating the coefficient of the threshold filter will be described later with reference to FIG. 4, and then back to FIG. 2.

The entropy encoder 50 entropy encodes the critically filtered image according to a probability of a symbol in the image to generate a compressed image. In a typical JPEG2000 standard, entropy encoding means is divided into Tier-1 and Tier2, and Tier-1 performs a bit plane on a code block and performs arithmetic encoding by extracting a context for each coding pass. Tier-2 assigns to each code block the cut off points using ratio distortion optimization. The entropy encoding unit 50 of FIG. 2 according to an embodiment of the present invention may appropriately use such conventional entropy encoding means according to an implementation environment and a necessity.

FIG. 4 is a block diagram illustrating a method of training a critical filter in a lossless image compression apparatus according to an embodiment of the present invention. do.

Assuming a diagnostic medical image as an input image, an N frame 12-bit diagnostic medical image is provided. The 12-bit image is converted into 8 bits, which is a range value of visually important data by the preprocessor 10, and band-divided into 13 bands by using the band division encoding by the converter 20.

In step 410, the average ( μ _i ) and the standard procedure ( σ _i ) are calculated for the divided 1 to 13 bands (ie, the band-divided i th frame).

)과 표준편차(

)를 산출한다.Next, in step 420, the average of the N frame bands from the mean and the standard deviation calculated in step 410 (

) And standard deviation (

) Is calculated.

단계에서는 최적

, 그리고

(r=odd) 단계에서는 최적

,

(r=even) 단계에서는

를 점진적으로 산출한다.In operation 430, the threshold parameter k _i is determined for each divided band by referring to a preset threshold table in consideration of visual loss of the input image due to compression.

Optimal for stage

, And

optimal at (r = odd)

,

(r = even) step

Calculate gradually.

)과 표준편차(

), 그리고 430 단계를 통해 결정된 임계 파라미터(k _i )를 이용하여 임계 필터링 계수(thr _i )를 산출한다. 이 때, 임계 필터링 계수는 앞서 설명한 수학식 1에 따른다. 즉, 420 단계를 통해 산출된 N 프레임의 평균에 N 프레임의 표준편차와 430 단계를 통해 결정된 임계 파라미터의 곱을 가산함으로써 임계 필터의 계수를 산출할 수 있다. 이렇게 산출된 임계 필터의 계수는 임계 필터링부(30)에 전달되며, 수학식 2를 통해 진단 의료 영상을 임계 필터링한다.In step 440, the average of the N frame bands calculated in step 420

) And standard deviation (

And a threshold filtering coefficient thr _i using the threshold parameter k _i determined in step 430. At this time, the threshold filtering coefficient is in accordance with Equation 1 described above. That is, the coefficient of the threshold filter may be calculated by adding the product of the standard deviation of the N frames and the threshold parameter determined in step 430 to the average of the N frames calculated in step 420. The calculated coefficient of the threshold filter is transmitted to the threshold filtering unit 30, and critical filtering the diagnostic medical image through the equation (2).

In operation 450, the reconstructed image calculated through the inverse transform unit 37 may be transferred to operation 460. In addition, the 8-bit image converted by the preprocessor 10 is also transmitted in step 460.

와

에 의해 임계 필터링으로 발생하는

(band Mean Square Error)와 임계 필터링 후, 역대역분할을 통해 얻는

(frame Mean Square Error)를 다음의 수학식 3과 같이 비교한다.In step 460, an error according to reverse band division may be obtained from the two received images.

Wow

Caused by threshold filtering by

(band mean square error) and threshold filtering

(frame Mean Square Error) is compared as in Equation 3 below.

단계를 제외한

단계의

의 초기 값은

단계에서 결정된 최적

이며, 470 단계의 최적화 판단 결과 상기 수학식 3을 만족하지 못할 경우 최적

이 결정될 때까지 N 프레임마다 다음의 수학식 4와 같이 점진적으로 증가되어 임계 필터링(30)을 다시 수행하게 된다.At this time,

Except steps

Phase

The initial value of is

Optimal determined in phase

If, as a result of the optimization determination of step 470 does not satisfy the above equation (3)

Until this is determined, it is gradually increased every N frames as shown in Equation 4 below to perform the threshold filtering 30 again.

는 i 번째 대역의

번째 스테이지 내의 k 번째 임계 파라미터이고, 실수 α는 임계 파라미터의 판정 기준점을 조절하는 임의의 변수로서, 예를 들어

과 같이 주어질 수 있다.here,

Of the i th band

Kth threshold parameter in the first stage, real α is any variable that adjusts the decision reference point of the threshold parameter, for example

Can be given as

In summary, the coefficient of the threshold filter is optimized by selectively updating the threshold table by comparing the band error according to the coefficient of the threshold filter with the error according to the reverse band division after the threshold filtering based on the pixel value of the divided band. In addition, the process of selectively updating the threshold table is achieved by gradually increasing the threshold parameter by a predetermined value every N frames if the two results do not match.

FIG. 5 is a flowchart illustrating a method of designing a threshold filter in the lossless image compression apparatus of FIG. 4 according to an embodiment of the present invention, which corresponds to the design process of the threshold filter described with reference to FIG. 4.

In operation 410, the input image of the frame N (N is a positive integer) is divided into a plurality of bands, and an average and a standard deviation of each band are calculated.

Next, in step 420, the average and standard deviation of the N frames are calculated from the average and the standard deviation calculated in step 410.

In operation 430, the threshold parameter is determined for each divided band by referring to a preset threshold table in consideration of visual loss of the input image due to compression.

In step 440, the coefficient of the threshold filter is calculated using the average and standard deviation of the N frames calculated in step 420 and the threshold parameter determined in step 430. This threshold filter is calculated by adding the product of the standard deviation of the N frames and the threshold parameter determined in step 430 to the average of the N frames calculated in step 420.

In operation 470, the threshold table is selectively updated by comparing the band error according to the coefficient of the threshold filter calculated in operation 440 with the error according to the reverse band division after the threshold filtering, based on the pixel values of the divided band. At this time, if both errors do not match as a result of the comparison, the threshold parameter is gradually increased by a predetermined value every N frames.

6 is a flowchart illustrating a series of performing operations for optimizing the threshold parameters of FIGS. 4 and 5, in which a medical diagnostic image is first read out, an average and a standard deviation for each divided band, and an average of N frames. And standard deviation.

Then, stage 0 is performed, the threshold parameter value is changed, and the coefficient of the threshold filter is calculated. In addition, the threshold parameters are controlled by comparing BMSE and TMSE. Subsequent stage 1 is similar to stage 0 but repeats the operation with different optimal threshold parameters. Each detailed execution step corresponds to the threshold filter setting method of FIGS. 4 to 5 described above, and a detailed description thereof will be omitted.

FIG. 7 is a block diagram illustrating an image compression device that guarantees a diagnostic lossless method according to another embodiment of the present invention, and corresponds to a configuration in which a quantization unit 40 is added to the image compression device of FIG. Therefore, a different configuration will be described with reference to the newly added quantization unit 40.

)를 통해 수행된다. 스텝 사이즈는 다음의 수학식 5와 같이 정의된다.The quantization unit 40 quantizes the critical filtered image through the threshold filtering unit 30 using the quantization step size calculated from the statistical information of the frequency band signal of the input image and the band weight. That is, if visual lossless compression is possible through the critical filtering unit 30, the diagnostic lossless compression may be achieved through the quantization unit 40. Such quantizers may be specifically termed visual quantizers, which are diagnostics obtained from the statistical characteristics (which can be mean and standard deviation) and band weights of frequency band signals of diagnostic medical laboratory images. Lossless quantization step size (

Is performed through The step size is defined as in Equation 5 below.

A more specific quantization process will be described below with reference to FIG. 8. FIG. 8 is a block diagram illustrating a quantization process in the image compression apparatus of FIG. 7 according to another embodiment of the present invention in detail. Here, quantization of a diagnostic medical image in which visual loss is guaranteed through the threshold filtering unit 30 is illustrated. Explain the center.

) 및 표준편차(

)를 가져온다. 또한, 820 단계에서 양자화부(40)는 대역 별로 최적화된 임계 파라미터(

)를 가져온다. 이와 더불어, 830 단계에서 양자화부(40)는 다음의 수학식 6에 정의된 바와 같이 대역 가중치(

)를 결정한다.In step 810, the quantization unit 40 performs an average of the band-divided N frames (

) And standard deviation (

). In addition, in step 820, the quantization unit 40 optimizes the threshold parameter for each band (

). In addition, in step 830, the quantization unit 40 may use the band weights as defined in Equation 6 below.

).

}이 된다. 이러한 대역 가중치는 부대역 영상의 에지(edge)에 대한 압축 효과를 고려하여 미리 설정된 것으로 에지 영역에 대한 압축률이 높은 것이 바람직하다.In this case, the weight values belonging to the subband layer n are {

} Becomes These band weights are preset in consideration of the compression effect on the edge of the subband image, and preferably have a high compression ratio for the edge region.

) 및 표준편차(

), 최적화된 임계 파라미터(

), 대역 가중치(

)를 기준으로 양자화 스텝 사이즈(

)를 결정하여 양자화한다. 양자화 스텝 사이즈는 앞서 설명한 수학식 5에 의해 산출된다. 즉, 양자화 스텝 사이즈는 임계 파라미터, 대역 가중치 및 N 프레임의 표준편차의 제곱을 승산함으로써 산출된다.Finally, in step 840, the quantization unit 40 calculates the average (

) And standard deviation (

), Optimized threshold parameters (

), Band weights (

) Quantization step size (

) To quantize. The quantization step size is calculated by Equation 5 described above. That is, the quantization step size is calculated by multiplying the threshold parameter, the band weight and the square of the standard deviation of the N frames.

FIG. 9 is a diagram for describing a quantization method in an image compression device that guarantees diagnostic lossless according to another embodiment of the present invention, and illustrates the quantization method of step 840 of FIG. 8.

)와 임의의 실수인 양자화 스케일(scale)(d)를 적용하여 양자화한다. 보다 구체적으로, 부대역 화소 값을 양자화 스케일 값 및 양자화 스텝 사이즈의 곱으로 제산한 값을 소수점 이하 버림함으로써 임계 필터링된 영상을 양자화한다. 이러한 과정을 다음의 수학식 7과 같이 표현될 수 있다.The quantization method uses a quantization step size (for a high frequency band band-divided by the quantization unit 40).

) And any real quantization scale ( d ). More specifically, the threshold-filtered image is quantized by rounding down the subband pixel value by the product of the quantization scale value and the quantization step size. This process can be expressed as Equation 7 below.

According to various embodiments of the present invention described above, the visual loss is guaranteed through a threshold filter calculated by using statistical information and a threshold table of subbands divided from an input image, and using a quantization step size based on band weights. By quantizing the image, it is possible to provide a high quality diagnostic image while improving the compression ratio of the image having a large data volume such as a diagnostic medical image.

On the other hand, the embodiments of the present invention described above can be implemented in a computer-readable code on a computer-readable recording medium. The computer-readable recording medium includes all kinds of recording devices in which data that can be read by a computer system is stored.

Examples of computer-readable recording media include ROM, RAM, CD-ROM, magnetic tape, floppy disks, optical data storage devices, and the like, which may also be implemented in the form of carrier waves (for example, transmission over the Internet). Include. The computer readable recording medium can also be distributed over network coupled computer systems so that the computer readable code is stored and executed in a distributed fashion. In addition, functional programs, codes, and code segments for implementing the present invention can be easily deduced by programmers skilled in the art to which the present invention belongs.

The present invention has been described above with reference to various embodiments thereof. Those skilled in the art will understand that the present invention can be implemented in a modified form without departing from the essential features of the present invention. Therefore, the disclosed embodiments should be considered in an illustrative rather than a restrictive sense. The scope of the present invention is shown in the claims rather than the foregoing description, and all differences within the scope will be construed as being included in the present invention.

10: preprocessor 20: converter
30: threshold filtering unit 33: threshold table
37: inverse transform unit
40: quantization unit 50: entropy coding unit

Claims (16)

In the design method of the threshold filter for lossless image compression,
Dividing an input image of an N (N is a positive integer) frame into a plurality of bands and calculating an average and a standard deviation for each band;
Calculating an average and standard deviation of N frames from the calculated average and standard deviation;
Determining a threshold parameter for each divided band by referring to a preset threshold table in view of visual loss of the input image due to compression;
Calculating a coefficient of a threshold filter by adding a product of the calculated standard deviation and the determined threshold parameter to the average of the calculated N frames; And
And selectively updating the threshold table by comparing a band error according to the calculated coefficient of the threshold filter with an error according to reverse band division after threshold filtering based on pixel values of the divided band. The method of claim 1,
And selectively updating the threshold table, incrementally increasing the threshold parameter by a predetermined value every N frames when the comparison results do not match the two errors. delete In a lossless video compression method,
Pre-processing the input image by performing DC-level shifting and tiling;
Band dividing the preprocessed image into a frequency band;
Performing threshold filtering by selectively outputting the pixel value by comparing a pixel value belonging to the divided frequency sub-band with a coefficient of a threshold filter; And
Generating a compressed image by entropy encoding the critical filtered image according to a probability of a symbol in the image,
The coefficient of the threshold filter is configured to divide an input image of an N (N is a positive integer) frame into a plurality of bands, calculate an average and a standard deviation for each band, and calculate an N frame from the calculated average and standard deviation. Calculates an average and a standard deviation, determines a threshold parameter for each divided band by referring to a preset threshold table in consideration of visual loss of the input image due to compression, and calculates the average and standard deviation of the calculated N frames Calculated by performing an operation using the determined threshold parameter. The method of claim 4, wherein
The threshold filtering step,
Compare pixel values belonging to the divided frequency subbands with coefficients of a threshold filter,
And when the pixel value is smaller than the coefficient of the threshold filter as a result of the comparison, 0 is output; otherwise, the pixel value is output as it is. The method of claim 4, wherein
The coefficient of the threshold filter is selectively updated by comparing the band error according to the calculated coefficient of the threshold filter with the error according to the reverse band division after threshold filtering based on the pixel value of the divided band. Characterized in that it is optimized. The method according to claim 6,
Selectively updating the threshold table, incrementally increasing the threshold parameter by a predetermined value every N frames if the result of the comparison does not match the error. The method of claim 4, wherein
And quantizing the threshold filtered image using quantization step size calculated from statistical information of a frequency band signal of the input image and a band weight. The method of claim 8,
The statistical information is the mean and standard deviation of the N frames,
The quantization step size is a value calculated by multiplying a predetermined band weight and a square of the standard deviation of the N frame in consideration of the calculated threshold parameter, the compression effect on the edge of the subband image. Way. The method of claim 8,
The quantizing may include quantizing the threshold filtered image based on a value obtained by dividing the subband pixel value by a product of a quantization scale value and the quantization step size. A computer-readable recording medium having recorded thereon a program for executing the method of any one of claims 1, 2 and 4 to 10. A lossless video compression device,
A preprocessor preprocessing the input image by performing DC-level shifting and tiling;
A converting unit converting the preprocessed image into a frequency band and band-split it;
A threshold filtering unit configured to compare the pixel values belonging to the divided frequency subbands with coefficients of a threshold filter and selectively output the pixel values by performing critical filtering; And
An entropy encoder configured to generate a compressed image by entropy encoding the critical filtered image according to a probability of a symbol in the image,
The coefficient of the threshold filter is configured to divide an input image of an N (N is a positive integer) frame into a plurality of bands, calculate an average and a standard deviation for each band, and calculate an N frame from the calculated average and standard deviation. Calculates an average and a standard deviation, determines a threshold parameter for each divided band by referring to a preset threshold table in consideration of visual loss of the input image due to compression, and calculates the average and standard deviation of the calculated N frames And calculating by using the determined threshold parameter. The method of claim 12,
The threshold filtering unit,
Compare pixel values belonging to the divided frequency subbands with coefficients of a threshold filter,
And outputting 0 when the pixel value is smaller than the coefficient of the threshold filter as a result of the comparison, and outputting the pixel value as it is otherwise. The method of claim 12,
The coefficient of the threshold filter is selectively updated by comparing the band error according to the calculated coefficient of the threshold filter with the error according to the reverse band division after threshold filtering based on the pixel value of the divided band. Device characterized in that it is optimized. The method of claim 12,
And a quantization unit configured to quantize the critical filtered image using quantization step sizes calculated from statistical information of the frequency band signal of the input image and band weights. The method of claim 12,
The statistical information is the mean and standard deviation of the N frames,
And the quantization step size is a value calculated by multiplying a predetermined band weight and a square of the standard deviation of the N frame in consideration of the calculated threshold parameter, the compression effect on the edge of the subband image.

Images