The Research Of Parallel Image Denoising Based On Nonlinear Diffusion Equation

Large-scale computing and data processing are required in image processing. At present, the digital image processing speed can not meet the requirements of real-time. To improve this situation, parallel computing technique was introduced into digital image processing. Parallel computing technology came from traditional serial computational technology, and would become the mainstream of the future calculation model. Imagery processing's parallelization design is in order to improve image processing speed, processing more image data in the limited space and time. Therefore, parallel image processing algorithm research is a currently very important and valuable direction.This research is mainly based on the nonlinear diffusion parallel algorithm for image denoising models. Firstly, make improvement of the nonlinear diffusion denoising model. The improved nonlinear diffusion denoising model with control function was introduced in view of the characteristic human visual masking effects. The improved model used the noise visibility function to construct a ? -direction diffusion controlled function, and used the mask brightness function to construct a ? -direction diffusion controlled function. Meanwhile, a stable numerical calculation solution was chosen for the improved model, and the simulation was achieved. Experimental comparison showed that the improved model was better in denoising, edge-preserving, details-keeping, and also improved the image visual effect. Secondly, make parallel design for the improved nonlinear diffusion denoising model. According to the shortcomings of the improved nonlinear diffusion denoising model such as intensive calculation, time consuming, etc, based on the built cluster for image processing platform, the parallel algorithm was designed starting from the partition strategy selection, load balance, image edge data calculation, super step division and several other aspects, and then realized the simulation. Experimental data analysis showed that the parallel algorithm of the denoising model can basically obtain a considerable same processing quality and visual effect comparing with the traditional serial denoising algorithm, and also a shorter running time, better acceleration ratio, higher parallel efficiency, which embodied the necessity and advantages of the parallel algorithm. Finally, make a further optimization of the parallel nonlinear diffusion denoising algorithm. Using the parallel hybrid model, a multi-level parallel structure was designed, and the simulation was realized. Experimental data analysis showed that the optimized parallel denoising algorithm was able to keep the same image processing quality as the previous parallel denoising algorithm, and parallel performance was better, verified the superiority of multi-level parallel algorithm.