Research On Block-matching And 3D-filtering Based On The Wavelet Transform

The image has gradually occupied the main position in the information transmission.The importance of image makes the research of digital image processing technology a hot topic,and the technology of digital image denoising occupies a large proportion in the research of digital image processing.In order to achieve a good image denoising effect,the research on image denoising has never stopped,and many outstanding achievements have been achieved.In many denoising algorithms,the non-local denoising algorithm has opened up a road for people and enriched the horizon.In many denoising algorithms,the non-local denoising algorithm not only has good denoising effect,but also has important significance.In the non-local denoising algorithm,the Bloc-matching and 3D-filtering algorithm combines with the knowledge of communication,which has a good denoising effect,and it also provides new ideas for people.However,there are serious defects in the traditional wavelet transform,which affects the processing effect of the algorithm.In this paper,The Bloc-matching and 3D-filtering algorithm is the main research content.In the research of the algorithm,we found that the traditional wavelets transform used in Bloc-matching and 3D-filtering algorithm in the defect of insufficient translation invariance and direction selectivity is poor.These defects will affect the algorithm performance,how to solve these problems is the focus of research in the paper.Aiming at the shortcomings of the algorithm,we propose to construct another wavelet transformation to sample the image information repeatedly,so as to achieve the improvement of approximate translation invariance and directional selectivity.In the implementation of the hypothesis,we have studied the wavelet transform deeply,we found that the characteristics of dual tree complex wavelet transform conform to the conditions we need.Therefore,in the improvement research of Bloc-matching and 3D-filtering,we decided to use dual tree complex wavelet transform to solve the problems of wavelet transform to achieve algorithm improvement.But in the learning of dual tree complex wavelet,it is found that the construction of traditional dual tree complex wavelet will lead to excessive complexity,though it can improve the effect of algorithm,but it will also cause the increase of denoising time.Therefore,for the construction of dual tree complex wavelet transform,we propose a fast construction method using Hilbert transform and analytic function.The theoretical algorithm is still need to be proved by practice.In the experiment,we choose the subjective method of eye observation and objective method of peak signal-to-noise ratio as the evaluation standard of the denoising algorithm's execution effect.The experimental results show that the improved algorithm has better denoising effect compared with the original algorithm and the traditional denoising algorithm.Finally,we summarize the work of the full text at the end of the paper,and make a prospect for the further direction of the improvement of the algorithm.