摘要
Image matting is inherently an ill-posed problem of determining the mixing parameter, "alpha", for each pixel in an "unknown" area of an image. Although there were many approaches designed to solve this problem successfully, their performance decreases rapidly both in accuracy and speed when matting the images with complex texture or color due to poor choices of the sample pairs of foreground and background. In this paper, an AIS (Artificial Immune System) algorithm IFEN (Immune Feature Extraction Network) is used to improve the quality of sample pairs by extracting "feature" foreground/background pairs from the original randomly sampled pairs. The experimental results show that the performance of matting is promoted significantly by our method.
- 出版日期2012
- 单位华南师范大学