| With the rapid development of communication technology, the color imagewhich was used as human receiving extemal information carrier becomes the bestform of expression of the real world with its vivid images. A color image is composedby RGB three components, so its data amount is three times of that gray image. Thestorage and transmission of digital images has become the key to the image encodingtechnique. The purpose of image processing not only improves the visual effect of theimage, but also improves the compression ratio as much as possible in ensuring in thesubjective visual unaffected, which means that reduces the amount of the data to savethe memory space as much as possible.The image source coding can achieve the purpose of saving the transmissionbandwidth or saving the storage space. Transform coding has become the mainmethod of the information-source coding. The basic approach is that we divided theimage into a certain size of the sub-picture block and then transformed them withorthogonal transform and finally processed the coefficient matrix obtained to achievethe purpose of compression. Transform is a lossy compression encoding. Both thecompression ratio and the compression time generated through the differentconversion mode were different. The correlation of the image between the coefficientsafter orthogonal transform is reduced. Most of the energy of the image is focused ononly a few coefficients. And using the appropriate quantization and scan codes and soon can effectively compress the data of image. The discrete cosine transform (DCT) isbest transform coding method under the minimum mean square error criterionbecause its eigenvectors is very close to the the base vectors of covariance matrix ofthe natural image. So it has a wide range of applications in image compression codinginternational standards.Taking into account three color components RGB of the color image from aphysical model, in addition to different colors, they have the same texture, edge andgray-scale changes in gradient, so they have strong correlation. However, the correlation between each color components has not been fully utilized until now, so itlimits the further improvement of compression performance. The multi-dimensionalvector matrix (MVM) theory can unify the space component and color componentsRGB of the color image information into a multi-dimensional matrix model. It can notonly remove the spatial redundancy of the image completely, but also remove thecorrelation among the color components by the orthogonal transform and improve thecompression efficiency of the encoder greatly. So it makes this problem as a piece ofcake.This paper is based on the multi-dimensional vector matrix theory, and provides avariable size segmentation quantitative algorithm. Taking into account the coefficientsfrom different areas of an image and difference size blocks have different statisticalproperties and different energy centralization. Then, the original color image isdivided into sub-three-dimensional matrix of a variable size according to the activitycharacteristics of the image. The variable size of the three-dimensional vector matrixof discrete cosine transform (3D-VMDCT) is applied to the corresponding size of thesub-three-dimensional matrix. We process the coefficients after DCT with using thethree-dimensional balanced nonlinear scalar quantization and entropy coding.Finally, this paper used Visual C++6.0as a platform, and programmed toachieve the quantization algorithm based on variable size segmentation ofmulti-dimensional matrix. Then we discussed the compression performance underquantify the conditions of different size blocks. The experimental results show that thecompression efficiency of the algorithm provided is superior to that of JPEG standard.The provided algorithm improved the signal-to-noise ratio at least2dB compared tothe JPEG standard, and also improved the quality of the reconstructed image. It hasverified the effectiveness and feasibility of the proposed method. |
PDF Full Text Request