Study On Random Noise Removal Method For Continuous Brain Image Sequences Based On Bilateral Filtering

Image denoising is an important branch of image processing.The image is affected by various factors during acquisition and transmission process,resulting in the generated image often containing noise.These noise not only cause the subjective visual quality to deteriorate,but also interfer the accuracy of subsequent image processing steps.The requirement for image denoising is to remove the noise introduced in the generation process while preserving the information of the image itself.Among a variety of denoising algorithms,the bilateral filtering is widely used for with its excellent ability of edge preserving.For the three-dimensional continuous brain image sequence data,the foreground signals have strong correlation between layers.The three-dimensional bilateral filter can filter the data in three-dimensional space,thus using this correlation to achieve better filtering quality than the conventional bilateral filterIn this paper,we propose a filter for three-dimensional data denoising named three-dimensional bilateral filter.The new algorithm directly filters the three-dimensional brain image as a whole,and makes full use of its interlayer correlation to further improve the filtering quality.In order to quantitatively analyze the denoising effect of the filter,we use the two indexes of image quality analysis,PSNR and SSIM.Aiming at the shortcomings of the time efficiency of the new algorithm,we analyze and compare several fast algorithms of bilateral filtering,and apply the piecewise linearity fast algorithm to the three-dimensional bilateral filter.The experimental results show that the two indexes of the three-dimensional algorithm are higher than the bilateral filter with the same parameters,which proves that the fast three-dimensional bilateral filter can effectively improve the filtering quality.Comparing the running time,the time efficiency of fast three-dimensional filtering is also more excellent.The denoising algorithm proposed in this paper extends the filtered space from two-dimensional to three-dimensional.Appling the proposed algorithm to the three-dimensional brain image,both the denoising effect and the time efficiency can be improved.The algorithm can be used for other types of three-dimensional data.