一种医学图像去噪程序的并行优化

目的：在采集、处理和传输过程中，医学图像会存在各种噪声，严重影响医学图像的质量和后续对图像的各种处理，因此医学图像去噪具有重要意义。同时医学图像数据量大，去噪处理算法复杂，在一般个人电脑上进行医学图像去噪仍是一个非常耗时的过程．很难满足实际应用中高实时性的要求．因此需要通过优化来提高去噪的处理速度。方法：本文利用CUDA（Compute Unified Device Architecture）并行编程对基于同质算法的三维医学图像去噪进行加速，CPU和GPU（Graphic ProcessorUnit）异构编程方式能发挥GPU高强度的计算能力，提高算法的执行速度。通过使用纹理存储器将图像数据与纹理绑定，优化存储器访问，提高数据访问速度。优化过程中，合理选择三维图像数据的分块方式和线程块维度。可以获得更快的加速。结果：与基于同质的matlab和CPU去噪程序相比，并行优化后GPU程序在保持去噪图像质量的前提卞可以达到几百倍的加速。结论：CUDA加速大大缩短了三维医学图像去噪的运行时间，解决了医学图像去噪的速度瓶颈问题．可以应用于对运行速度有要求的图像处理中。

Parallel Optimization of a Medical Image Denoising Program

Objective： In the process of acquisition, processing and transmission, medical images are easily infected with noises, which seriously affect the quality of medical images and subsequent image processing, so medical image denoising is an important research subject. With massive data and complex algorithms, medical image denoising is a time-consuming process in personal computers. Also it hardly meets the demands of application requirements, so it＇s necessary to optimize to obtain high speed.Methods： Using the CUDA （Compute Unified Device Architecture） programming to speed up the three-dimensional medical image denoising based on homogeneity similarity. The heterogeneous programming of CPU and GPU （General Purpose Unit） provides strong compute ability and improves the speed. Using the texture memory and binding texture to the image data, optimizes the memory access, speeds up the data access. Meanwhile, choosing the reasonable method of image data partition, obtains a larger speed-up. Results： Compared to the matlab version of denoising program as well as the CPU version, the GPU achieves a speed-up of hundreds while maintains the quality of the denoised medical image. Conclusions： CUDA programming shortens the executing time of three-dimensional image denoising largely, solves the speed bottleneck of medical image denoising, and it can be applied to applications which demand high speed.