A MLP-PNN Neural Network for CCD Image Super-Resolution in Wavelet Packet Domain

Image super-resolution methods process an input image sequence of a scene to obtain a still image with increased resolution. Classical approaches to this problem involve complex iterative minimization procedures, typically with high computational costs. In this paper is proposed a novel algorithm for super-resolution that enables a substantial decrease in computer load. First, decompose and reconstruct the image by wavelet packet. Before constructing the image, use neural network in place of other rebuilding method to reconstruct the coefficients in the wavelet packet domain. Second, probabilistic neural network architecture is used to perform a scattered-point interpolation of the image sequence data in the wavelet packet domain. The network kernel function is optimally determined for this problem by a MLP-PNN (Multi Layer Perceptron - Probabilistic Neural Network) trained on synthetic data. Network parameters dependent on the sequence noise level. This super-sampled image is spatially filtered to correct finite pixel size effects, to yield the final high-resolution estimate. This method can decrease the calculation cost and get perfect PSNR. Results are presented, showing the quality of the proposed method.