Vector Median Filters Based on Noise Detection for Color Images

A novel filtering method is proposed in this paper in order to improve sharpness and detail reserve ability of color image filter. Start with analyzing the correlation between neighboring pixels and the correlation between each image channel, the image is then processed with a preprocessing procedure. Then we adopt one step singular correlation method to detect noise and the noise been detected is then complemented using the most relevant vector filtering. The simulation results show that this method we proposed can not only effectively detect the pepper and salt noise but also effective in reservation and restore of image edge and details. The PSNR and MSE value of the proposed method are significantly improved.