CUDA下局部阈值和二值函数的边缘检测

随着GPU编程的发展,海量数据并行运算更加普及,GPU编程已演化为高度并行、多线程、多核处理的高速大数据计算能力,具有较高的存储器带宽。局部阈值和二值函数下的边缘检测是将3×3图像通过局部阈值函数转换到位块中,而位块中出现的一些模式则被定义为边缘检测算子,该算法能够更快地对图像进行有效的边缘检测。文中介绍了局部阈值和二值函数边缘检测原理,以及如何利用CUDA实现检测,并通过调试仿真将传统边缘检测与局部阈值和二值函数边缘检测的时间进行对比,以证明局部阈值和二值函数边缘检测更快更有效。

Local Threshold and Boolean Function Edge Detection Algorithm Using CUDA

Edge detection is an important subject in image processing due to its importance in a wide range of applications. Massive parallelization of algorithms has become very popular in the recent years with the development of the programmable. Graphic Processor Unit, or GPU, has evolved into a highly parallel, multithreaded, manycore process or with tremendous computational power and very high memory bandwidth. A completely different concept is used in the local threshold and Boolean function based edge detection. In that case 3 x 3 blocks of the image are converted in bit blocks. Some patterns appearing in the bit blocks are considered as indicators of edge existence. This article introduces the principle of local threshold and Boolean function edge detection in detail and its implementation by using CUDA. This algorithm proves more effective and faster than traditional edge detection by comparing the time consumed.