Image Denoising By Fractal Non-local Algorithm

This paper focus on an important issues of Image denoising, that’s How to remove the noise and also keep and other important information. For example, edge and texture. But most of present denoising algorithm is based on the average of the methods to achieve the effect of denoising. That’s because the method of average can’t correctly looking for similar points in texture and near the edge. It will lead to the image over-smooth or image edge blur.Through deeply research of fractal related theory, we believe that different fractal dimension will correspond to different areas of the image. These areas including, smooth area, the edge area and texture area, etc. That’s because the grey value of smooth area changes small, it have smaller fractal dimension; Edge regions due to the influence of the gray changes, it have larger fractal dimension; And texture area are have relatively fixed fractal dimension. So, fractal can help us search for the right similar points in different areas of the image. So that, We can get a more satisfactory image denoising results.Based on the above theory research and analysis, we put forward a new image denoising algorithm, The Fractal local mean algorithm, it improve the existing mainstream denoising algorithm, the local mean denoising algorithm. The main method is to calculate the fractal dimension of each point by box counting dimension. And then, it combines the fractal dimension with the Local mean to search for the similar points. Finally, we use the correct method to looking for the similar points mean, get better denoising results. Through a series of validation and contrast test, we can verify the new algorithm is feasible. At the same time, through the calculation of different image denoising PSNR (peak signal to noise ratio, PSNR), It further proof that the new algorithm for image denoising can get more satisfying de-noising result.