Impulse noise removal based on SVM classification

This paper presents a new impulse noise removal technique based on Support Vector Machine (SVM). When an image is affected by impulse noise, removal of impulse noise is required so that the quality of the image is enhanced. Based on the SVM classification, the entire pixel in the test image is assigned to a particular class (`0´ or `1´) where `0´ is designated as noisy pixel and `1´ is designated as non-noisy pixel. If the central pixel is found to be corrupted, it is assigned by `0´ and accordingly median filtering will be processed. In other case, if the central pixel is assigned with `1´, pixel will be unchanged. This proposed technique for removal of impulse noise from gray images is superior to some of the techniques available in the literature. The peak signal to noise ratio and structural similarity show how this method outperforms some of the existing techniques in this field.