摘要
提出了一种基于图形处理硬件的递进式直线段检测方法,该方法利用图形处理硬件实现了边缘提取、霍夫变换、端点检测等步骤。其中的GPU并行霍夫变换采用θ参数空间及图像空间双重划分的方法,并使用线程块内的共享存储器较好地解决了同一θ空间的各线程同步问题。实验表明,该方法既有递进式检测方法的准确性,又充分利用图形处理器强大的计算能力获得了较高的实时性。
An iterative line segment detection procedure by graphic hardware was presented, which implemented edge detection, Hough transform, end points detection with GPU. The GPU based parallel Hough transform was designed for both θ parameter space segmentation and image space segmentation, and threads for the computation of the same θ parameter space were properly synchronized by shared memory in thread block. Experimental results prove the accuracy of the method for its iterative property, and good real-time capability because of full utilization of GPU's powerful computation effort.
出处
《计算机应用》
CSCD
北大核心
2009年第5期1359-1361,共3页
journal of Computer Applications
基金
广西省自然科学基金资助项目(0731020)
关键词
图形处理器
计算统一设备架构
并行霍夫变换
Graphics Processing Unit (GPU)
Compute Unified Device Architecture (CUDA)
parallel Hough transform