Denoising Of Non-local Method

Along with the gradual development of the information age and the popularity of image processing, image quality has been increasingly and highly demanded. Image is often corrupted by noise in its collection, acquisition or transmission. The noise is the main factor that influenced image quality and seriously affected to extract the information from the image. So, it is required to remove the noise from the image before analyzing and using the image, and image denoising is one of the widely discussed topics in computer image processing and computer vision.Non-Local Means is an emerging and effective image denoising method. The NL-Means tries to use the natural image's periodic case and its high degree of redundancy. It searches for similar grey blocks in the whole image, and the estimated value is computed as a weighted average of all the similar pixels in the image. As we know, it only considers the pixel's grey property but ignores the geometric attributes. In order to sufficiently use the self-similarity of the image's local geometric structures, we propose three NL-Means image denoising algorithm based on structure tensor for similarity measure, which is capable of describing the self-similarity of local geometric structures in the image. They are as follows:(1) Improved Non-Local Means Algorithm based on structure tensor: using "block andblock" matching strategy.(2)Improved Non-Local Means Algorithm based on structure tensor: using "block andpoint" matching strategy.(3) Rotationally invariant block matching strategy improving Non-Local Means denoising method.The experimental results demonstrate the effectiveness of our algorithm, and also show that our method has advantages for both edge preserving and PSNR value.