一种自适应非极大值抑制边缘检测算法及FPGA设计实现  

Adaptive Non-maxima Suppression Edge Detection Algorithm and FPGA Design to Achieve

作　　者：王文涛[1] 朱顺安[1] 唐菀[1] 

机构地区：[1]中南民族大学计算机科学学院,武汉430074

出　　处：《小型微型计算机系统》2014年第8期1926-1930,共5页Journal of Chinese Computer Systems

基　　金：国家自然科学基金青年基金项目(61103248)资助

摘　　要：边缘检测是机器视觉和计算机视觉处理的基础性工具,其效率和效果直接影响了后续图像处理的质量及视觉系统的整体响应时间.用FPGA平台固化边缘检测算法可以大大提高算法的效率.Sobel算法计算简单,但其检测出来的边缘粗,同时现有FPGA平台上采用的固定域值、经验阈值的实现方案通用性较差.为此,本文采用非极大值抑制方法来改进Sobel算子,并运用一种基于梯度直方图的二阶导数最大值的阈值选取策略使算法具自适应性.在FPGA平台上,采用流水线及分布式策略设计出边缘检测算法的并行运算电路.仿真结果表明与现有FPGA平台上的Canny与Sobel算法相比,改进后的边缘检测方案在保证实时性的前提下,提高了边缘检测的质量,具备了自适应性、FPGA门资源占用少等特性.Edge detection is a fundamental tool in machine vision and computer vision. Its efficiency and effects will impact on the quality of subsequent image processing and the response time of visual system. Edge detection algorithm in FPGA can greatly improve the efficiency. Sobel algorithm is simple, but the rough edges detected. Meanwhile, using a fixed threshold or experience threshold based on FPGA platform is less versatile. In this paper, we come up with an improved Sobel algorithm based on the non-maxima sup- pression edge detection and apply a new threshold selection strategy based on the max second derivative method of gradient histogram to make the algorithm to be more self-adaptive. In the end, a parallel computing circuit is designed by using the pipelines and distrib- uted strategies in FPGA. The simulation results, comparing with Canny and Sobel operation in FPGA, indicate that the proposed edge detection algorithm improves the quality of the edge detection without losing real-time.

关 键 词：边缘检测 梯度 非极大值抑制 FPGA 

分 类 号：P391[天文地球—地球物理学]