Modified Image Restoration Algorithm Using Neural Network Based on Harmonic Model
Yang Bing, Zhang Zhi-gang, Shufen Dai, Xiao Zhongnan · AASRI Procedia · 2012
Abstract In this paper, we considered the neural network image restoration algorithm based on harmonic model and improved the efficiency of the algorithm. After analyzing the complexity of the algorithm, I found that the storage of the connection weight wastes a lot of spaces. And when I made a deeper analysis of the, I found that there are many duplicated elements, and the distribution of the elements is regular. So I started from this point to improve the algorithm, and simplified the calculation of the connection weights matrix, and avoided a large amount of data storage. As a result, the space complexity reduced to from, meanwhile, the time complexity is improved. This is a great improvement. At last, I made some data simulation experiment, and the result give a further description of the algorithm.