Dominant Color Detection and Grayscale Propagation Based Color2gray Method

As state-of-the-art color2 gray methods can not guarantee the simultaneous preservation of global and local contrast, as well as the values of grayscale pixels, a color2 gray method based on dominant color detection and grayscale propagation is put forward in this paper. Firstly, this paper proposes a histogram peaks based dominant color detection method to extract a small representative colors set for the input image. Secondly, these dominant colors are mapped to grayscale space with gradient domain based color2 gray method, which can preserve the contrast in color space as much as possible. Then, grayscale values of dominant colors and grayscale pixel are propagated into nearby colors through local linear embedding model. Finally, this method detects if there are pixels which haven’t been decolorized and transform them to grayscale ones. Numerous experimental results show that this method simultaneously preserves the global and local contrast as well as the grayscale pixels and thus the diverse scene as shown in the input color image.