摘要
常用的视频分割算法包括光流法、背景减除法和帧差法,但是3种算法容易受到计算量、环境噪声以及亮度突变等因素的影响,造成分割效果不理想。为了改善视频分割的质量,能够较好效果的提取到运动物体,提出一种将边缘检测与帧差法相融合的算法。该算法同时在两个方向进行。一是对相邻两帧图像进行边缘检测,将图像的边缘进行差分以得到运动物体的轮廓,为了连接轮廓中可能会出现的细小断点,对差分结果进行闭运算。二是根据设定的阈值,将相邻两帧图像的对应像素值转换为二值化数据,对二值化结果进行差分。最后综合闭运算与二值化差分结果,当数据发生突变时,即认为是运动物体部分。实验表明,该融合算法可以实时的检测到视频中的运动物体,效果较好。
Optical flow method, background subtraction and frame difference is commonly used in video segmentation algorithm. However, the amount of calculation, environmental noise and brightness mutations likely to affect the three algorithms, Resulting segmentation is not ideal. In order to improve the quality of video segmentation, and can have a good effect to extract a moving object, Proposed an algorithm fusion frame difference and edge detection. The algorithm simultaneously in two directions. One is to detect the edges of two adjacent frames, difference the object edge and close the result. Two is change the corresponding pixels value of two adjacent frames into two binary data based on threshold, difference the binary. Finally, the mutation data is the object part according to the close operation and two value difference results. Experiments show that the fusion algorithm can be detected moving object in real-time, and the effect is good.
出处
《电子设计工程》
2015年第12期109-112,共4页
Electronic Design Engineering
基金
河北省高等学校创新团队领军人才培育计划(LJRC003)
关键词
边缘检测
帧间差分
形态学
视频分割
FPGA
FPGA
edge detection
frame difference
morphology
video segmentation
FPGA
