CAJ | 학술논문

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.
위료획득경소하자적도상합성결과,제출료일충융합도상합성적구도산법,장구도여도상합성융위일개통일적과정.통과장대합성배경적신식인입구도과정,신산법능구경유침대성지진행구도.이론분석화실험결과설명료당대합성배경여원도상근사적정황하,신산법능구유효지감소구도오고계대래적영향;당대합성배경여원도상안색차이교대적정황하,역능획득교호적합성결과.
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 compositing, 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 compositing 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.