Research On Multiple Images Denoising Based On Wavelet Transformation And BM3D

With the rapid development of information, people pay more and more attention on image and use it as data carrier for its easy to transport and large amount of information. Currently, an image contains information that people want but also a lot of noise information because of that image acquisition is affected by environment and internal circuit of image-acquisition instruments. Noise information may disturb the results of subsequent image-processing, such as image mosaic, integration, recognition and so on. Image denoising is the premise of image processing, and denoising effect affects the subsequent image-processing directly. Wavelet transformation has developed for a long time to become relatively mature and it has been applied in many areas. At the same time, denoising method based on non-local information has a very good development in recent years, typically Block Matching and 3-D Filtering method(BM3D), which has a good denoising effect in the aspects of perception and PSNR(Peak Signal Noise Ratio).Through research and analysis, multiple images denoising method is proposed combined wavelet transformation with BM3 D based on that several images taken under the same condition have more primitive information, that is BM4 D algorithm. Efficiency of image denoising is improved by means of algorithm optimization to reduce computing budget. Firstly, one of multiple images is denoised through wavelet transformation, then the first step of BM3 D is used to group similar blocks. Group results are applied to the other images, and blocks in the same group are processed, so that there is no increase in the amount of computing similarity matching blocks, but also imports more real information of the original image to improve the denoising effect. Secondly, blocks are grouped roughly by comparing the gray value, and then grouped accurately by computing the Euclidean distance. Comparison results are recorded to reduce the number of data comparison for improving efficiency of the algorithm. Finally, effect and execution time of the aboved denoising algorithm are compared by simulation. The experiment results show that perception and PSNR and efficiency of this method have been improved considerably. Summary of experiment results provides a direction to optimize algorithm, which has great significance and value.