Edge Detection of Noise Image Completely Based on ICA
Hongyan Chen, Ling Yang, Shang Ma · International Conference on Electric Information and Control Engineering · 2012
The classical methods are very sensitive to noise when used to edge detection of noise image. Because of the problem, a new method of edge detection for noise image completely based on Independent Component Analysis (ICA) is proposed in the paper. First, sparse coding shrinkage algorithm based on independent component analysis (ICA-SCS) is used to reduce noise component, and then we can obtain a processed image and other two images by proper geometric transformation of the processed image. These three images are strongly correlative. Then sharpening image can be obtained by analysis the matrix information of the three images by means of ICA. The experimental results show that the method can extract edge and texture information accurately even of image corrupted seriously by Gaussian Noise.