Application of improved Quasi-Newton method to the massive image denoising

Nowadays, sensors have generated more and more images than before, and parallel processing capacity shows great importance in massive image denoising tasks. Those images are always various in quality and hard to recognize by human or computer. In consequence, massive image denoising are essential. In this paper, two kinds of filter based on Quasi-Newton method called CB and BB filter are proposed, which takes the Newton iteration algorithm as the mathematical basis. Both two filters were achieved with MATLAB to produce the n product n matrix filter. The difference is that the CB filter process the pixels from the image center to the edge, while the BB filter process the pixels from the upper left of the image to the lower right boundary. To illustrate the effectiveness of CB and BB filter, we analyze key indicators after the massive image denoising with massive remote sensing image and high resolution image. We also compared the two filters with the traditional FastICA algorithm. The results indicate that the CB and BB filter have their own advantages in different type of image. The two filters both can effectively improve the massive image quality and enhance the visual effect.