Color Image 3-D Matrix Transform Compression Coding Based On YC Submatrix

Along with science and technology development and progress, The colorimage is moved into computer world from real world. The key technique whichmakes the move come true is the color image compression. The reason is that ifthe color image isn't compressed, it takes up many memmories. So it would bedifficult to transmit and storage a lot of color images in computer world. But thecolor image compressed just takes up lesser memories, which makes it possible totransmit and store many color images.The reason that the color image can be compressed is that the color imagecontains lots of redundancies from the viewpoint of information theory. We can'tknow the rule for a color image only by eyes. But image data has redundancies inthe viewpoint of statistics. For example, pertinence exists among pixels andpertinence exists among components of one pixel. All of these are representationways of image redundancy. Generally speaking, the redundancies of the colorimage include statistic redundancy, structure redundancy, knowledge redundancy,vision redundancy and color space redundancy, etc. Since Oliver put forward PCM coding theory in1948, data compressiontechnique has developed fast. And there are about one hundred methods now,such as, Huffman coding, arithmetic coding, REL, PCM , DCT, wavelet coding,etc. Any of these methods has better effect to wipe off some kind of redundancy.But the effect isn't good if we use only one method to compress image. Thereason is that there are many kinds of redundancy in one color image. So weoften adopt several coding methods to compress the color image. And thesuccessful example is JPEG standard. The main content of this paper is on the color image compression. Butdifferent from prevenient article, this paper is to compress the color image in theidea of 3-D matrix. The thought of 3-D matrix has been brought forward for shortof time. And one kind of the 3-D matrix transform method has already been putforward;that is WDCT. But it still need research to how to apply thistransformation to compress image better. WDCT was used on the basis of RGB3-D sub-matrix before. Then the transformed result was quantified by VQ. Thecompression ratio is high in this way. But this way is complex and is relative withtraining image because of using VQ. So in this paper scalar quantization is used.We need adapt color space and the way of sub-matrix partition to scalarquantization in the process of the color image compression using 3-D matrix.If we choose RGB color space to cooperate with scale quantization, thecompression effect isn't good enough. In view of human eyes permitting someextent error for the color image, we choose YCbCr color space. This space issuitable for the color image compression. Because Y component of this space ismore important than Cb and Cr components unlike RGB space where everycomponent is the same important, we can sample Cb and Cr components. This isthe purpose of choosing this space. Because of choosing the different color space,the method of 3-D sub-matrix partition is also different. YC sub-matrix partitionmethod is used in this paper, and we put forward the concept of Y class 3-Dsub-matrix and C class 3-D sub-matrix. This partition method can exert betterpredominance of 3-D matrix WDCT to wipe off the redundancy of components.In this paper quantization method is linear uneven scalar quantization. Thisquantization way is based on human eye's characteristic. That is to say that subtlequantization is adopted for the areas where human eyes are sensitive and coarsequantization is adopted for the other areas where human eyes aren't sensitive.Both Huffman coding and arithmetic coding are used in this paper. WhenHuffman is used, Huffman code-table lookup is adopted. The Huffman code-tableis result from lots of statistic experimentations. And prior to Huffman coding, theRLE and DPCM are used, in order to reach better compression effect. Thearithmetic coding is duality arithmetic coding.At last this paper makes experiments using Visual C ++ 6.0 on the Windowssystem. And it also compares this technique with basic JPEG and 3-D matrixtransform based on VQ. The compare with basic JPEG shows that the techniqueof this paper is a little better than basic JPEG method. And the compare with 3-Dmatrix transform based on VQ shows that this technique is better on the aspect ofreal time application and common capability. So the compare result shows thatcompression effect of this paper technique gets to some extent achievement.Although this paper is on the color image compression based on 3-D matrix,the results have some meaning for future research. On one hand, the result ofexperiments shows that 3-D matrix transform together with scalar quantizationcan result in high compression ratio. Although the predominance isn't very big,this technique can be regarded as the basis for further research. On the other hand,the technique of this paper is based on 3-D matrix, and it can be boosted to 4-Dmatrix for video compression and multi-D matrix for other things by furtherresearch.