Impulse Noise Removal from Digital Natural Images in a Large Range of Noise Density Based on Adaptive Mean and Median Filtering

In this paper, a new hybrid algorithm has been presented for identifying and eliminating impulse noise in digital images. The main idea of this algorithm is the detection of the noise percentage of the image, and selection of two different methods for removing the noise in high and low densities. The difference between the two methods of noise removal is in their idea for selecting most appropriate window size. In both cases, after determining the appropriate size, the proposed idea is to replace the window’s central pixel with the average amount of median and mean of neighboring non-noisy pixels. In the proposed algorithm, only the noisy pixels of the image are processed, and the noise free pixels are not changed. The ideas applied in this paper lead to desirable performance in all densities of noise. In addition, ignoring the processing of non-noisy pixels and choosing proper processing method in high and low densities of noise reduces the computational cost. Experimental results on standard test images represent that the proposed algorithm has a better performance in terms of visual quality and PSNR (Percent of Signal to Noise Ratio) quantitative measure, compared to other proposed algorithms.