Weighted Median Filtering of Im Based on Grads Similarity

Through analyzing existing methods of median filtering on salt & pepper noise, a new improved weighted median filtering based on grads similarity is proposed. The proposed algorithm divides the image with salt & pepper noise into uncertain noise pixel and signal pixel from the difference of pixel grey value in filtering window. The uncertain noise pixels are further analyzed by grads similarity in filtering window. The disturbance of noise pixel is removed by applying extremum trimming method. At last, a improved weighted median filtering algorithm with abandon points of max square deviation is used to compute the grey value of noisy point. Simulation results show that this proposed method has a great improvement in removing salt & pepper noise and preserving the details of image. The restored image can also obtain better peak signal to noise ratio (PSNR).