Neural Network And Svm-based Image Compression (coding) Theory And Methods

SVM(Support Vector Machine) is the statistical learning theory of classification and regression, which put forward by V.Vapnik of AT & T Bell Laboratories, and it is a Special neural network. It is shown that SVM is an excellent method in classification (recognition) and data modeling fitting in practice.This paper is based on the National Natural Science Foundation project on "artificial brain information processing new neural network model" No.60673 101 for the research tasks, the author applies the data coding thinking based on virtual source modeling which is put forward by her instructor, first introduces SVM(support vector machine) into data compression field. By using two modeling to BP neural network and SVM source, the author put forward the image coding (compressed) Program based on SVM. In addition to, we establish a SVM image compression experiment system, and analyses the components and functions in the system, give some examples of image compression. Experiments show that the system can get a higher compression ratio for image and the restored image has a better visual effect.