Image Denoising Based On Low Complexity Non-Local Means And Kalman Filtering

With the rapid development of the information era, it has become more convenient to exchange and transmit information than before. Exchanging Image transmission is a direct and convenient approach among the diversified approaches of information transmission. A noise-free image can enable its user to pick out the information immediately. Unfortunately, in image transmission, noise disturbance to image could not be completely avoided, which may be caused by different intrinsic (i.e., transmission equipment) and/or extrinsic (i.e., environment) conditions. A noise-contaminated image will not only reduce the accuracy of the image, but also make post-processing difficult. Hence, image De-noising becomes an important part of image processing to acquire image information.This thesis proposes a method for image De-noising using low complexity Non-Local Means and Kalman filtering based on the several traditional methods of noise reduction. In this thesis, the main achievements and innovations are as follows:1. A method of image De-noising based on MCMC sampling of the non-local means image noise reduction method is proposed. An in-depth study of the traditional non-local means image noise reduction method which has several drawbacks such as low efficiency, how to choose the weighting function, how to selected similar images, how to accelerate it processing speed, while maintaining or improving noise performance due to some deficiencies, is carried out in this thesis. The proposed method as compared with the traditional method of non-local means image noise reduction method has an improvement in noise reduction efficiency and processing speed. From the simulation results, the proposed method outperforms the traditional non-local means noise method and other common noise reduction methods. Using the traditional non-local means noise reduction method the image smoothing degree is larger after processing, but proposed method greatly reduces the image smoothing degree after processing In addition, the computational complexity of proposed method in contrast to the traditional method of non-local means with similar other methods for de-noising is relatively low. Again, from simulation results, and was found to achieve state-of-the-art De-noising performance in terms of both peak signal-to-noise ratio (PSNR) and mean structural similarity (SSIM) metrics when compared to the other methods.2. In this thesis, Kalman filter is used to reduce noise, by introducing the concepts of the state space and variables and recursive algorithm, the proposed method have solved the inhibition problem which can’t be solved by others, and it’s applicable to the stationary process and non-stationary process, which resolves limiting factors of other methods. Simulation result shows that the Kalman filtering method can remarkably reduce the noise of the original images, and solve the fuzzy problem associated with filtering. Compared with other traditional methods, this algorithm can deal with the noise much better, and keep the information such as the lines, points and the details of the edge of the image almost without any loss. Furthermore, NSHP model is used to reduce the amount of computation, obtain the changes in the correlation of the information of the images, so this model can describe the images easily and grasp the main features of the images. In addition, in using the NSHP model to update the current region of the pixels, we just consider the pixels nearby and ignore the impact of the pixels outside a certain range in order to obtain more accurate useful and practical information that can greatly reduce the amount of computation in the Kalman filter to update.3. Finally, a De-noising system interface based on MATLAB/GUI is designed in the last Chapter. First of all, we drew a sketch of the interface to determine the buttons and function menus and then designed the interface according to the requirements of the function based on the sketch. We then wrote callback procedures of the menus and buttons. Finally, we tested the function of each module the results of each test showed us the effectiveness of each module. Going through all the above steps, we completed the design of the interface of the De-noising system; effectiveness of the system was validated by experimental results.