摘要
串行Sobel梯度算子边缘检测算法需要将两个掩模S1和S2分别在图像的每个像素上移动,并在每个像素上进行11次加法运算,即需要11×N2次加法,时间复杂度为O(N2);文章提出了一种Sobel算子分解模型,并设计了一种在SIMD-MPP模型上基于Sobel分解算子的并行图像边缘检测算法,该并行算法总共只需要8次平移操作和9次加法运算即可完成,其时间复杂度为O(1),加速比达到N2,大大地提高了基于Sobel算子的图像边缘特征提取的效率。
The image edge extraction on the basis of sequence Sobel gradient operator must move two mask S1 and S2 on every pixel of image and add 11 times, i.e., the Sobel operator requires addition 1xN^2 times. The time complexity is O （N^2）. In this paper, a decomposing model of Sobel operator is proposed, and a parallel algorithm of image edge extraction for Sobel decomposing model based on SIMD-MPP model is designed. The parallel algorithm only requires data movement 8 times and addition 9 times. The Time complexity is O（1）. The speedup of the parallel algorithm is N2. Thus, the efficiency of image edge extraction based on Sobel operator promote greatly.
出处
《微电子学与计算机》
CSCD
北大核心
2006年第9期132-134,共3页
Microelectronics & Computer
基金
国家自然科学基金项目(60572080)
