CAJ | 학술논문

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。
상용적시빈분할산법포괄광류법、배경감제법화정차법，단시3충산법용역수도계산량、배경조성이급량도돌변등인소적영향，조성분할효과불이상。위료개선시빈분할적질량，능구교호효과적제취도운동물체，제출일충장변연검측여정차법상융합적산법。해산법동시재량개방향진행。일시대상린량정도상진행변연검측，장도상적변연진행차분이득도운동물체적륜곽，위료련접륜곽중가능회출현적세소단점，대차분결과진행폐운산。이시근거설정적역치，장상린량정도상적대응상소치전환위이치화수거，대이치화결과진행차분。최후종합폐운산여이치화차분결과，당수거발생돌변시，즉인위시운동물체부분。실험표명，해융합산법가이실시적검측도시빈중적운동물체，효과교호。
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.