基于感知颜色空间的自然图像抠图

Perceptual Color Space Based Natural Image Matting

提出了一种基于感知颜色空间的快速自然图像抠图算法 该算法首先手工把图像进行区域划分;接着采用一个简单实用的模型估计前景和背景颜色;最后提出了新的基于感知颜色空间的透明度估计算法,把透明度的计算细分为亮度和色度透明度的计算与已有的算法相比,在具有同等效果的情况下。

This paper proposes a fast natural image matting method based on perceptual color space. This method begins from segmenting regions by a user, and then uses a practical model to estimate the foreground and background color components of each unknown point. Last, we present a novel perception based alpha estimating approach by estimating chroma and intensity alpha respectively. This method is much faster than the existing approaches of comparable quality.