Cluster Based Implementation of Non Local Means filter for Denoising of Textural patterns
Digital Image Processing forms an important element in the generation as well as transformation of an image. As far as an image is concerned, its quality and clarity matters. To restore the images, there already exists several image filters belonging to both local and non-local categories. Each of these filters have their own advantages and drawbacks especially in terms of clarity and computational speed. To overcome these issues, we propose an efficient non-local means filter for the restoration of the image based on arranging the data in the form of a cluster tree. Structuring of the data is done on the basis of preselection of patches and we make use of distance as the parameter while patch selection. Hence, this approach contributes to larger speedups, better quality and less computational cost especially when the filter is truly non-local.