An edge-preserving image reconstruction using neural network

This paper presents an image restoration model based on the implicit functi on theorem and edge-preserving regularization. We then apply the model on t he subband-coded images using the artificial neural network. The edge infor mation is extracted from the source image as a priori nowledge to recover t he details and reduce the ringing artifact of the subband-coded image. The multilayer perceptron model is employed to implement the restoration proces s. The main merit of the presented approach is that the neural network mode l is massively parallel with strong robustness for the transmission noise a nd parameter or structure perturbation, and it can be realized by VLSI tech nologies for real-time applications. To evaluate the performance of the pro posed approach, a comparative study with the set partitioning in hierarchic al tree (SPIHT) has been made by using a set of gray-scale digital images. The experimental results showed that the proposed approach could result in compatible performances compared with SPIHT on both objective and subjectiv e quality for lower compression ratio subband coded image.