Saliency Based Decolorization

With the rapid development of technology, color digital media plays an important role in our daily life. However, grayscale is still one of the most popular pictorial forms, widely used in newspaper, e-book, black-and-white printer. This paper presents a new decolorization approach using saliency including color and position information to maintain the original contrast.From informatics perspective, image contains information. Color image is consist of three channels, contains more information compared with grayscale, which is consist of only one channel, thus decolorization is a dimension reduction process. Conventional methods of decolorization only use color difference without considering the spatial distribution of color, therefore bring more information loss and lead to contrast reduction.Saliency is used to represent image in terms of visual attention, containing both color and space information. Important areas with high saliency value should still be noticeable, or could be argued to remain saliency value, after decolorization. Highlighted with this observation, we propose to add saliency to the decolorization process and use an optimization approach to preserve saliency.Our main contribution is to apply saliency theory to decolorization process by three steps:(1) Propose a simplified contrast model(SCM) which denotes pixels’ level of noticeability in terms of color difference and spatial relationship.(2) Put forward an energy function for the purpose of minimizing the gap between color image’s saliency and corresponding grayscale’s saliency with weak color order using SCM, where grayscale is represented by a mapping function of chromaticity variation.(3) Accelerate saliency calculation by loosening or strengthening constraint using matrix calculation.The experiment results show that our methods are better than state of art decolorization methods in terms of preserving global contrast distribution of grayscale and color image.