An Improved Median Filter Based on Efficient Noise Detection for High Quality Image Restoration

To restore images corrupted by impulse noise, various image filtering strategies have been proposed. However, a common drawback of these methods is that the details and edges can not recovered satisfactorily when the noise level is high. A novel method based on efficient noise detection algorithm is proposed in this paper for effectively denoising extremely corrupted images and better details and edges preservation. The proposed method replaces only noisy pixel by either the median value or by a noise-free pixel value inside the selected filtering window. While uncorrupted pixels are remain unchanged. The key feature of the proposed method is very high noise detection accuracy, yielding zero miss-detection rate. Comparative studies of the proposed method and other methods for different images are made and the advantages of the proposed method are fully demonstrated.