New image super-resolution scheme based on residual error restoration by neural networks
Fengzhi Pan · Optical Engineering · 2003
The scheme proposed combines an existing image interpola- tion algorithm with an artificial neural network (ANN) used to model the residual errors between the interpolated image and the respective origi- nal image. Mathematical analysis shows that the performance of the proposed method is superior to that of known single-frame interpolation algorithms. The image restoration results using the presented scheme indicate that the restored images are very similar to the real high- resolution images. We also illustrate that the performance of any single- frame interpolation algorithm can be enhanced by combining the inter- polation algorithm into our scheme. Experimental results show the proposed method on generalization and computation complexity is su- perior to other neural network schemes. © 2003 Society of Photo-Optical Instru- mentation Engineers. (DOI: 10.1117/1.1604397)