A JPEGLS Based LosslessLossy Compression Method for TwoDimensional

A JPEG-LS Based Lossless/Lossy Compression Method for Two-Dimensional Electrophoresis Images 1 Source: 2003 International Conference on Informatics, Cybernetics, and Systems Authors: Kevin I-J Ho, Tung-Shou Chen, Hui-Fang Tsai, Mingli Hsieh, and Chia-Chun Wu Speaker: Chia-Chun Wu (吳佳駿) Date: 2004/12/09

Outline v Introduction v Schema v Compression Method v Decompression Method v Results v Conclusion 2 NCHU

Introduction 3 v We use Lossless and Near-Lossless compress the important areas and unimportant areas in Two-Dimensional Electrophoresis (2 D-Gel) images. v Our system improves traditional JPEG-LS to enhancing the compressed image quality. NCHU

Schema (1/2) v Compression flow chart 3 Original 2 D-Gel Image JPEG-LS Near-Lossless Compress 4 1 Near-Lossless Compressed File Detect Protein’s Areas 2 ５ 4 ５ Record Boolean Value of Important Areas Write Difference Value of Original Image’s Important Areas ６ Difference value Record File NCHU

Schema (2/2) v Decompression flow chart Near-Lossless Compressed File Difference value Record File １ ３ JPEG-LS Near-Lossless Decompress Add Difference to Image’s Pixel Value ４ ３ ２ Near-Lossless Decompressed 2 D -Gel Image 5 NCHU Keep Important Information of 2 D -Gel Image

Compression Method (1/5) v Original 2 D-Gel image － This is an original 2 D-Gel image. X-axis represented the PH value of protein and Y-axis represented the amount Y-axis molecular weight. 6 73 72 75 78 71 71 75 18 4 74 76 5 15 16 10 73 18 23 28 74 75 74 10 73 77 X-axis NCHU Fig. 1 Original 2 D-Gel image

Compression Method (2/5) v Fetching protein’s area. －To collect the important protein ‘s areas of 2 D-Gel image. The colourful areas will treat as important areas, and white areas will treat as unimportant areas. 7 0 0 0 0 18 4 0 0 5 15 16 10 0 18 23 28 0 0 0 10 0 0 Fig. 2 The important part of 2 D-Gel image NCHU

Compression Method (3/5) v Transform the important parts to Boolean value － Boolean value True(1) represents important areas, whereas False (0) represents unimportant areas. 8 0 0 0 0 1 1 0 1 1 1 0 0 0 1 0 0 NCHU Fig. 3 Boolean value record file of important part

Compression Method (4/5) v Image after JPEG-LS Near -Lossless compression －This is an 2 D-Gel image after traditional JPEG-LS Near-Lossless compression. 9 71 73 77 76 74 72 74 17 6 73 79 3 18 13 8 73 21 23 26 74 74 77 12 75 76 Fig. 4 Image after JPEG-LS Near-Lossless compression NCHU

Compression Method (5/5) v Difference value records important part － The difference value of 2 DGel image via the original image and lossless compression will store in a record file. 10 0 0 0 1 -2 0 0 2 -3 3 2 0 -3 0 2 0 0 0 -2 0 0 Fig. 5 Difference value record file NCHU

Compression Example 73 72 75 78 71 71 75 18 76 5 4 74 15 16 10 71 73 77 76 74 - 72 74 17 79 3 6 73 18 13 8 = 0 0 0 0 1 -2 0 0 2 -3 3 2 73 18 23 28 74 73 21 23 26 74 0 -3 0 2 0 75 74 10 73 77 74 77 12 75 76 0 0 -2 0 0 Original 2 D-Gel Image after JPEG-LS Near-Lossless compression Difference value record file 11 NCHU

Decompression Method (1/3) v The image of Near-Lossless decompression － This is a decompressed image after traditional JPEG-LS Near. Lossless compression. 12 71 73 77 76 74 72 74 17 6 73 79 3 18 13 8 73 21 23 26 74 74 77 12 75 76 Fig 6. Image after JPEG-LS Near-Lossless NCHU compression

Decompression Method (2/3) v Modify the protein’s area of important part － Next, we base on the difference value of pixels for modifying the protein’s area of important parts. 13 0 0 0 0 1 -2 0 0 2 -3 3 2 0 -3 0 2 0 0 0 -2 0 0 Fig. 7 Difference value record file NCHU

Decompression Method (3/3) v The lossless image of our system’s important areas － This complete 2 D-Gel image is though traditional JPEG-LS Near-Lossless compression technique. 14 71 73 77 76 74 72 74 18 4 73 79 5 15 16 10 73 18 23 28 74 74 77 10 75 76 Fig. 8 Our system’s lossless compression NCHU of important part.

Decompression Example 71 73 77 76 74 0 0 0 0 1 -2 0 0 2 -3 3 2 73 21 23 26 74 0 -3 0 2 0 73 18 23 28 74 74 77 12 75 76 0 0 -2 0 0 74 77 10 75 76 72 74 17 79 3 6 73 18 13 8 Image after JPEG-LS Near-Lossless compression 15 + Difference value record file NCHU 71 73 77 76 74 = 72 74 18 79 5 4 73 15 16 10 Our system’s lossless compression of important part

Results (1/4) v 16 Fig. 9 is the result of the amplification of dotted frame in Fig. 1. Fig. 9 Partial magnify image of original 2 D-Gel image NCHU

Results (2/4) v 17 Fig. 10 is the result of the amplification of dotted frame in Fig. 4. Fig. 10 Partial magnify image of traditional JPEG-LS. NCHU

Results (3/4) v 18 Fig. 11 is the result of the amplification of dotted frame in Fig. 8. Fig. 11 Partial magnify image of our system. NCHU

Results (4/4) Table 1. The Comparison of image quality in traditional JPEG-LS Near-Lossless with our system (PSNR value). Result (4/4) 19 NCHU Unit: d. B

Conclusion 20 v We store the unimportant areas by Near-Lossless method. But we store important areas by Lossless method. It is very important to medical images. v Under different lossless level, we can find out our system has better image quality than traditional JPEG-LS. v Therefore, how to compress the size of record file and detect the protein’s location more correctly becoming an important topic in the future. NCHU