Digital Image Mixed Noise Filtering Algorithms

As an unpredictable random signal,noise,which is usually processed with the mathematical probability of statistical methods. Noise plays an important role and it has been highly evaluated in image processing, collection, it also has an important influence in handling all aspects of the output,specially for The image signal input and acquisition.Noise suppression is a critical problem.If the input image is polluted by a high intensity noise, which will definitely affect the whole process and the quality of the output image signal. A good image processing system is indeed needed, not only in the digital processing of computer but also in the analog processing, which takes noise reducing as the main target.Noise includes gaussian noise, impulse noise, additive noise and multiplicative noise.Mixed noise,composed of Gaussian noise and Impulse noise is a typical noise in digital images. Traditional noise filter methods can not filter mixed noise effectively causing the loss of image detail and edge.This thesis mainly studies the superposition of noise mixed with salt and pepper noise and Gaussian noise. The bilateral filter is a kind of edge-preserving denoising filter, which is determined by geometric spatial distance of filter coefficients and the decision of the pixel difference between the two functions of the filter coefficients. Image edge detection makes use of the differences of the object and the background of image features,such as grayscale, color, texture features, that is, to detect the location of the image characteristics where has been changed. Pulse coupled neural networks is a kind of imitation of biological neural computing systems, which simulates triggering of the optic nerve and feedback mechanism, which has a good performance in positioning mixed noise point. This project is based on the advantages of bilateral filter in filtering out the mixed noise, Inspired by the algorithm of PCNN in filtering mixed noise,the bilateral filter algorithm has been improved to select the most appropriate parameter values.In this paper, an optimization algorithm based on bilateral filtering and image edge detection has been proposed, according to texture direction distribution. The gray area of each quadrant-bit values is set as the criteria to identify noise points.The traditional bilateral filter algorithm is improved to filter Gaussian noise and Impulse noise recursively.Simulation results show that the algorithm has a strong ability in filtering mix noise, and a good performance in preserving edges and image details by using Matlab.