Research Of Compression Coding Algorithm For Video Based On Human Visual System

With the rapid development of information technology, multimedia communication plays an important role in the human life.Digital image and video which carrys a huge amount of information has been particularly widly applied.3G mobile communication technology is applied further,digital television and high-definition home theater are more and more popular.At the same time,digital cinema film gradually replace the traditional film,the performances of digital cameras and other electronic products have improved rapidly,visual information of image and video becomes more and more important in people's production and people's life.Meanwhile, the quality requirements of visual information have become more sophisticated,so has the image /video encoding techniques.The large amount of digital video and image data bring difficulties to its storing, transmitting and processing,so the technology of video and image encoding is the key technology.The main purpose of video and image encoding is to save storing capacity, reduce transmitting channel capacity and shorten the processing time.As a result of the development of computer,communications,microelectronic and network technology, video and image compression and transmission is possible,so people can enjoy the convenience brought about by technological progress.In this paper, we first introduced the video coding theory,such as Predictive Coding,Transform Coding and entropy coding. We also introduce international standards,for example,JPEG,H.26X,MPEG,and so on.These are classic image and video encoding method.During the development of image and video encoding,we should take full advantage of classical methods,and propose new methods.The main contents of this paper is the study of color image and video enconding. Unlike before,however,compressing the color image and video in this article is based on multi-dimensional vector matrix transformation theory.Traditional method of image compression is to transform RGB space to YCbCr space,this approach is to deal with image color component apart,rather than as a whole,which in some degree restricted color image compression.Because in RGB color space,each color component of color image has great relevance,for example,they have the same texture,edge and gray-scale changes in gradient,etc.Traditional compression methods can not remove the revelance.At the same time,there is also great relevance between the frame of video data.The compression methods based on the multi-dimensional orthogonal matrix vector algorithm in this paper is just to remove those relevance,that is,regard the RGB color space as a whole,and as for the video, regard eight consecutive video frame Y or U or V as a whole,then do a whole transformation.This method not only can reduce the correlation of internal components and between the data,but also be beneficial to reduce the time frame redundancy to achieve the purpose of further compression.We combined multi-dimensional vector matrix with discrete cosine orthogonal transform theory,to find the multi-dimensional vector orthogonal DCT matrix. Taking orthogonal transformation with multi-dimensional vector DCT orthogonal matrix on color image and video,the result of transformed data have a good focus on the distribution.For data distribution,we proposed a flexible three-dimensional scalar quantization method, which is taking use of the frequency region characteristics of transformed three-dimensional matrix, in view of the distribution of the image and video data, the scalar three-dimensional matrix approach is introduced. The role of quantization is to keep the subjective impression in a certain fidelity,and remove those information has less impact on the visual effects.Generally speaking,low-frequency component impact main aspects of image information,while high-frequency component impact the details of image.The coefficients near the origin of the DC coefficient take more contribution to the image, that need to ensure the quantitative accuracy of these coefficients,while coefficients farther from the origin, take the smaller contribution to image,that not need consider its accuracy in the quantization. When define quantitative matrix, the coefficient which is further away from the origin corresponds to a larger step quantization value,to make those coefficients zero. Removeing the frequency components which does not affect the visual effects, to achieve effective compression.Improving the compression effect and guarantee the quality of restored image as possible.In order to make zero coefficients centralized and convenient to copress,according to the distribution of the coefficient matrix,this article proposed two three-dimensional scanning ways, and maked use of the scanning ways to redistribute the data. The experiment results show, that after scanning the distribution of the data significantly better than pre-scan data,and the continuity of zero increasing. The non-vanishing data distribution is quite centralized,that is advantageous to enhance efficiency of compression.At last doing run length coding and entropy coding.Although multi-dimensional vector matrix theory is still in its early research stage of innovation, we know that this theory be applied to image and video encoding is valid.There are still many facts can be improved and developed for the technology.For example,it needs to go through a large number of experiments, to find out Huffman or arithmetic entropy coding method that match the characteristics of three-dimensional coefficients matrix;to take a flexible block and frame prediction mode for video,and then take multi-dimensional matrix vector transform coding; combine with the multi-dimensional vector matrix theory and the classical tradition methods,and also to play its unique advantages,the theory need to combine with AVS and H.264.More details need study and research for further.At last this paper makes experiments using Visual C ++ 6.0 on the Windows system. And it also compares this technique with basic JPEG. The comparation with basic JPEG shows that the technique of this paper is a little better than basic JPEG method. Through the validity of the experimental results, the multi-dimensional matrix transformation theory combined the multi-dimensional scalar quantification theory, is an advisable compression method, which is building a foundation for the further research.