Perceptual Segmentation Based Color2Gray

The conversions from color images to grayscale image are practically useful in digital image processing and can be used for image preprocessing and for gray-scaled printers. Generally, many important visual features will be lost in the grayscale results due to the lack of gray information from the color images. Both feature preserving and computational cost are the main concerns when designing a color-to-gray algorithm in the literature.First, we review and analyze some classical algorithms. All the algorithms are classified into two types: fast global techniques and gradient based methods. The fast global techniques use some functions to continuously transform the color space into gray space, and the gradient based methods use the gradient computed by pixel contrast to rebuild grayscale image. The fast global techniques often have faster speed while the gradient based methods have a better effect.We present a new grayscale transformation method based on image segmentation. As we know, human perception is mainly based on the segmentation of the image, so the algorithm based on segmentation may have a better visual result. Our algorithm use CIE Y channel information, and get the better result through the correction of the features lost in grayscale transformation. Through the segmentation with hue value of HSV color space, we can get the important visual feature lost in grayscale transformation. Then we can get the contrast of each patch and enhance the contraction of important feature. Finally, we constrict the adjustment of each patch to maintain the global gray effect. The result of our algorithm shows that our algorithm has a good effect and a boost of computation speed contrast with gradient based methods...