直方图均衡化结合中值滤波的CUDA并行化实现

Parallelization Implementation of Histogram Equalization Combined with Median Filtering Based on CUDA

直方图均衡化是图像增强技术中的重要算法。针对直方图均衡化不能有效去除噪声的问题,论文引入中值滤波算法来抑制最常见的椒盐噪声对直方图均衡化图像的干扰,并且利用GPU的并行计算能力,实现了基于CUDA和OpenCV的直方图均衡化结合中值滤波的并行化。实验结果表明,串并行处理图像效果一致,明显消除了图像噪声、提高了图像对比度和清晰度,但与传统CPU串行处理方式相比,对于分辨率为3216x2136的图像,获得了5.6倍的加速比,有效提高了数据处理速度。

Histogram equalization is an important algorithm in image enhancement technology.For the problem that histogram equalization can not effectively remove noise,the paper introduces median filtering algorithm to suppress the interference of the most common salt and pepper noise on histogram equalization image,and by the parallel computing power of GPU,parallelization implementation of histogram equalization combined with median filtering based on CUDA and OpenCV.The experimental results show that the serial parallel processing image has the same effect,which obviously eliminates the image noise and improves the image contrast and clarity.However,compared with the traditional CPU serial processing method,for images with a resolution of3216 x2136,an acceleration ratio of 5.6 times is obtained,which effectively improves the data processing speed.