摘要
为了获得更少瑕疵的图像合成结果,提出了一种融合图像合成的抠图算法,将抠图与图像合成融为一个统一的过程。通过将待合成背景的信息引入抠图过程,新算法能够更有针对性地进行抠图。理论分析和实验结果说明了当待合成背景与原图像近似的情况下,新算法能够有效地减少抠图误估计带来的影响;当待合成背景与原图像颜色差异较大的情况下,亦能获得较好的合成结果。
In order to achieve visually plausible image compositing results, this paper proposed an algorithm which integrated image matting and compositing into a unified process. By considering the attributes of the new background image for eomposiring, the new algorithm could extract the matte with more targeted information. Experimental results and analysis show that the algorithm can effectively reduce the artifacts in the eompositing result when the original image and the new background share similar colors and can also achieve good composing result when the colors of the original image and the new background are extremely different.
出处
《计算机应用研究》
CSCD
北大核心
2009年第12期4808-4810,共3页
Application Research of Computers
基金
国家自然科学基金资助项目(60573019)
广东省自然科学基金资助项目(05103541)
关键词
抠图
鲁棒采样
图像合成
随机行走
image matting
robust sampling
image compositing
random walks