KR20010097394A - method for different compression of the medical image - Google Patents

Abstract

In order to reduce storage space and improve transmission efficiency in PACS (Picture Archiving & Communication System) and telemedicine, the medical image which is lossless compression (DPCM) of diseased image and lossy compression (Fractal) image of diseased image in medical image A differential compression method of is disclosed. In the present invention, in the transmission of a medical image, a diseased image is recognized by a lossless compression using a DPCM compression algorithm, and a non-disease image is loss-compressed and transmitted using a fractal compression algorithm.

Description

Method for different compression of the medical image

The present invention relates to a differential compression method of a medical image, and in particular to the lossless compression (DPCM) of the image having a diseased part of the medical image to reduce the storage space and increase the transmission efficiency in PACS (Picture Archiving & Communication System) and telemedicine The disease-free image relates to a differential compression method of a lossy compression (Fractal) medical image.

In general, image compression is classified into lossy compression and lossless compression depending on whether image information is lost after reconstruction. Joint Photographic Experts Group (JPEG) and Moving Picture Experts Group (MPEG) techniques based on DCT, which are representative transform coding techniques, are lossy compression methods. However, medical imaging for diagnosing disease requires high image quality because of its particularity. Therefore, lossless compression is preferred over lossless compression with higher compression rates. The lossless compression method may be implemented using entropy coding techniques such as Huffman coding, Lampel-Ziv coding, and Arithmetic coding. In addition, the proposed algorithms for lossless compression of medical images include HINT (Hierarchical Interpolation), Difference Pyramid (DP), Bit-Plane encoding, and block coding. HINT is a variable resolution pyramid coding method based on sub-sampling. First, the resolution of the original image is reduced to 1/2 for each of horizontal and vertical, and the lowest resolution is encoded, and the reduced image is interpolated. It is a method of encoding a difference from an original image by doubling it by using (Interpolation). This method is approximate lossless since it interpolates four pixels of the upper pyramid using interpolation from a low-resolution pixel. DP is a compression method based on a variable resolution model such as HINT, which is composed of the average pyramid and the difference pyramid.

Therefore, when the medical image used in medical care is stored or transmitted without compression when using PACS or telemedicine, it takes up a lot of storage space and is not efficient to transmit. In addition, if the compressed and stored or transmitted medical image can reduce the storage space and the transmission rate is improved, there is a problem that the important information of the disease area is lost due to the nature of the medical image is affected a lot.

The present invention has been invented to solve the above problems, the disease image after loss of the disease image using the DPCM compression lossless compression after the disease image and the non-disease image to compress and store differential transmission and storage, non-disease The image is lossy compressed using a fractal compression method. Fractal compression method has high compression effect and eliminates deterioration and blocking of image at the boundary part, which is a problem of coding technique by DCT. The object of the present invention is to provide a differential compression method of medical images to reduce storage space and improve transmission rate.

1 is a block diagram illustrating an encoder and a decoder of a lossless predictive coding system for performing a differential compression method of medical images according to the present invention.

FIG. 2 is a diagram illustrating an example of a pixel used for a prediction value of a lossless prediction coding system for performing a differential compression method of a medical image according to the present invention.

3 is a flowchart illustrating a fractal compression method used for transmission of a non-disease image in the differential compression method of a medical image according to the present invention.

4 is a flowchart showing the restoration of fractal compression used for transmission of non-disease images in the differential compression method of medical images according to the present invention.

The present invention for achieving such an object,

In the transmission of medical images, a diseased image is recognized by the image lossless compression using the DPCM compression algorithm, the non- diseased image is transmitted by lossy compression using a fractal compression algorithm.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings.

1 is a block diagram illustrating an encoder and a decoder of a lossless prediction coding system for performing a differential compression method of a medical image according to the present invention, and FIG. 2 is a differential compression method of a medical image according to the present invention. FIG. 4 is a diagram illustrating an example of a pixel used for a prediction value of a lossless prediction coding system. 3 is a flowchart illustrating a fractal compression method used for the transmission of a non-disease image in the differential compression method of a medical image according to the present invention, and FIG. 4 is a non-disease image in the differential compression method of a medical image according to the present invention. This is a flowchart to show the restoration of fractal compression used for transmission.

In the present invention, the recognized image is divided into a disease image and a non- disease image, and the disease image is compressed by the DPCM algorithm, and the non- disease image is compressed by the fractal algorithm. Lossless compression of a medical image with a recognized diseased region is compressed by the DPCM method, which compresses by using the property that adjacent pixels have almost the same value.

A system for performing a differential compression method of a medical image includes an encoder and a decoder including the same predictor as shown in FIG. 1. Sequential every pixel of the input video Is input to the encoder, and the predictor generates the predicted value of the pixel from several past values. The output of the predictor is a near integer The difference or prediction error is as shown in <Equation 1>.

Equation 1

The prediction error shown in Equation 1 is encoded by using a variable length code in the encoder to make the next element of the compressed data stream. From the variable length code received Reconstruct and reverse the reaction as in <Equation 2>.

Equation 2

Various local, global, and adaptive methods Used to make In most cases, however, the prediction is formed by Equation 3 as a linear combination of m previous pixels.

Equation 3

In Equation 17, m is the order of the linear predictor, and rounding is a function used to represent a near integer. when i = 1,2,3, ..., m Is the prediction coefficient. In prescan applications, the subscript n indicates the predictor output according to the time of occurrence. In other cases n is used to represent the spatial coordinates or number of frames in the image. For example, in the first linear prediction encoding, <Equation 3> may be represented by <Equation 4>.

Equation 4

The variables imposed in Equations 1 through 4 are functions of spatial coordinates x and y. In DPCM, the difference between the predicted value X and the actual value x By encoding, the smaller this interval is, the closer to the actual value can be predicted. Usually, many pixels are used to predict a pixel value.

2 illustrates an example of a pixel using a predictive value. In FIG. 2, the sum of the pixel value B on the left side and the pixel value E on the upper side can be used as the value divided by two. Frequently used prediction methods are summarized as in <Equation 5>.

Equation 5

DCT-based coding of lossy coded images reduces high frequency components to increase the compression ratio, causing degradation of the image near the boundary, and causing blocking due to quantization of low frequency components. To overcome this drawback, a fractal compression algorithm has been developed that has excellent compression effect regardless of resolution. Fractal theory is based on self-similarity, in which parts of an object resemble the whole. The principle of image coding using fractals is based on the property that the image to be encoded using local self-similarity converges to a fixed point by repeated reduction change irrespective of the initial value. Instead it is encoded. Fractal coding has the advantage of better image quality than other coding methods at a low bit rate. A fractal structure with self-similarity refers to a structure made by repetition of a similar structure in itself, and an iterated function system refers to the self-similarity of an image by the coefficients of the Affine transform. It is summarized as in <Equation 6> as a system for reconstructing the ultimate image by repeatedly applying them from the image.

Equation 6

In <Equation 6> Denotes an Affine transform, and I is an arbitrary image. In particular, if the Affine transform is a reduced transform that satisfies <Equation 7>, the transform Repeatedly apply to convergence to a fixed point.

Equation 7

In <Equation 7> Represents the distance function At a reduction ratio to be. In general, when a fixed point exists for a given iteration function system, the fixed point is called an attractor. conversion The dragger is easily obtained when is given, but it is not simple to construct a conversion system having the same or similar dragger as the given image. What constitutes such a transformation system is a fractal encoding process. In fact, it is difficult to obtain an iterative function system that can represent the entire image. Therefore, in fractal coding, the iterative function system is applied by dividing the entire image. The Affine transform equation of the segmented iteration system applied to the compression of the gray image is shown in <Equation 8>.

Equation 8

In <Equation 8> Indicates the location, Represents the expansion, reduction, rotation, etc. of the Affine transform according to the combination, Moves the matching block, Is the gray level, The contrast scaling, Represents the difference in the average gray level. The iterative function system of the whole image represents the union of the fixed points obtained through each divisional iteration system. In real fractal coding, we first reduce the transform and then use Simplify In <Equation 8> Is fixed to 1, And Of the matching block Encodes the mean gray level difference between coordinates and matching blocks. Collage theorem provides the basis for the numerical answers of fractal coding. In other words, convert Let's say A's drag ratio Is transformed into a given image I and its own The smaller the difference from, the closer the image I is to the drag A.

Equation 9

Therefore, after constructing a divisional repetition function system for the image to be encoded, Affine transformation is performed for arbitrary initial values. If repeatedly applied, it is possible to reconstruct an image encoded by collage theorem.

Fractal compression used for the transmission of non-disease images first divides the image into non-overlapping region regions as shown in FIG. 3 (step S1). The region region covers the entire image regardless of size and shape. At the end of the initial segmentation, the area areas larger than the area areas that are necessary for the entire image cover and can be overlapped are collected (step S2). Closely match the area areas, and select the class of the Affine transform defined by the Affine coefficients appropriate for the area areas. (Step S3) When a compression process is performed to record a fractal image format composed of headers and clear selection information from the most significant region to the last region according to the list of Affine coefficients selected for each region region, Independent of the resolution of the image, a file with the identification parameters defined by the equation for the image is generated. (Steps S7-S8)

As shown in Fig. 4, the fractal reconstruction and the reconstruction process are symmetrical, and the reconstruction algorithm can reconstruct the fractal repetition (echo) using a given Affine coefficient similar to the compression process. have. That is, the unpacked Affine transform and region segmentation information are read from the fractal image generated during the compression process (step S10), a memory buffer is generated for the region and the region screen, and initialized (steps S11-S12). And it is an iterative process to relocate the data to each region region transformed from the legitimate region using the stored Affine coefficients to output the image of the last region.

<Table 1> shows the compression ratio of each image. When the diseased image is lossless compressed by DPCM, it is 1: 1.41 for gastric image, 1: 3.84 for chest image, 1: 2.64 for CT image. The ultrasound image was compressed to 1: 3.51 and vascular image to 1: 2.54. Also, the non-disease image is 1: 15.16 for the gastric image compressed by the fractal algorithm, which is a lossy compression method, the chest image is 1: 28.91, the CT image is 1: 5.38, the ultrasound image is 1: 31.62, and the blood vessel image. It is shown to be compressed to 1: 14.97.

The lossless compression shown in Table 1 shows an average compression ratio of 2.78 times, indicating no significant difference from the original image. In addition, the loss compression showed an excellent compression ratio of 19.21 times, but showed no blocking effect compared to the original image.

<Table 1> is compression ratio for each image

Image Chest CT image Ultrasound Blood vessel imaging Non-disease video: Lost compressed video 15.16: 1 28.91: 1 5.38: 1 31.62: 1 14.97: 1 Disease Image: Lossless Compression Image 1.41: 1 3.84: 1 2.64: 1 3.51: 1 2.54: 1

As described above, the disease image according to the present invention is lossless compression using DPCM compression, non-disease image loss loss compression using a fractal compression method can reduce the storage space without being affected by medical treatment and improve the transmission rate It is effective to let. Although the preferred embodiments of the present invention have been described in detail with reference to the accompanying drawings, the present invention is not limited thereto and may be improved or modified by those skilled in the art within the scope of the technical idea of the present invention.

Claims (2)

In the transmission of a medical image, a differential compression method of a medical image for lossless compression of the diseased image using the DPCM compression algorithm, and lossless compression of the diseased image using the fractal compression algorithm. The method of claim 1, wherein the fractal compression algorithm of the disease-free image comprises: first dividing the image into non-overlapping regions; Collecting the area areas larger than the area areas that are necessary for the entire image cover and are also overlapable; Selecting a class of Affine transforms that are close to the region region and defined by the Affine coefficients suitable for the region region; Recording a fractal image format consisting of headers with explicit selection information and headers from the most significant region to the last region according to the list of Affine coefficients selected for each region region to generate a file with the identification parameters defined by the formula for the image. Differential compression method of medical images, including.