The image intensity surface for the local neighborhood of image is well fitted by the least squares support vector machine (lssvm)

and then the gray level interpolation is implemented on the fitted intensity surface. The interpolation coefficient matrix of the local neighborhood is deduced from the lssvm with the radial basis function (rbf) kernel function

as an example. A method using the interpolation evaluating merit figure

psnr

to optimize the svm parameters is proposed. With the selected parameters

the computer interpolation experiments are carried out. The experimental results demonstrate the svm based interpolation algorithm has similar performance to cubic one but providing higher efficiency.