Multiscale wavelet support vector machine for image approximation
Hui Cheng, Chi Lu, Han Hai, Jinwen Tian · 2007
In this paper, a new multiscale wavelet support vector machines model (MWSVM) is proposed, according to the Mercer condition, the wavelet frame theory and kernel function nature. From the statistical learning theory and the SVM model, pointed out the SVM essence is kernel method, the different kernel function has decided the different SVM. The choice of kernel parameters is simplified in MWSVM. By the experiment with the single-variable two-variable function and real image, the new model can approach linear and the non-linear combination functions very well. The experimental result shows that MWSVM is the validity and the usability.