欢迎您！东篱公司
退出
-
申报数据库
-
立项数据库
-
成果数据库
-
项目获奖数据库
中文摘要：
视频序列图像的背景切换是从视频图像帧中自动准确的提取出对象图像，然后赋予对象图像新的背景．针对此问题，本文提出基于图割理论的图像对象提取算法．首先提出采用能量最小化模型分离图像对象背景，并利用最小割／最大流算法实现能量最小化模型．其次，为了提高算法分割效果，算法采用HSV颜色模型，削弱V分离的作用从而减弱亮度变化对于分割的影响；最后对分割后的对象二值图像进行修正，提出用形态学操作进行边缘修补，用目标同化进行空洞填补．实验证明，算法提取出的对象图像完整、图像边缘清晰，较好的实现了视频序列图像的背景切换．
英文摘要：
Object image extracted from video frame and compounded to a new background image, that is automatic switching video background. For this problem, a new algorithm based graph-cut method is proposed. First, energy minimization model is presented to separate object image and background image. And min-cut/max-flow method is used to implement this model. Then, in order to improve effects of our algorithm, HSV color space is used to increase results insensitive to illumination changing by weaken V tenn. Finally, revise the object binary image. Morphological operations method and target assimilation method are presented to repair the image edge and the error hole. Experimental results indicate that the outcome object image is satisfied and a new video segment can be created through switching background automatically using our algorithm.
同期刊论文项目
同项目期刊论文
期刊信息
位置：