Research On Parametric Image Decolorization With Local Contrast Enhancement

Decolorization is to transform the colorful picture into the grayscale image,and is just to reduce the dimensionality of the input image.The main task is to try to preserve the visual information of the original colorful picture and to make us recognize the image regions that are recognizable in the color image.It is widely used in the fields of image preprocessing,image printing and displaying.Therefore,it is worthy of studying and researching.In the previous decolorization methods,the color images are usually mapped into the grayscale image uniformly and automatically.Sometimes,we need to enhance or weaken the contrast depending on our subjective wishes.In order to highlight the foreground regions which attract the user's attention,this paper propose a parametric image decolorization algorithm with local contrast enhancement.The proposed method does not only support for choosing the interesting image parts with an interactive method but also support for obtaining the interesting parts automatically using the saliency detection to guide local contrast enhancement.Hence,we can control the contrast of each part of the grayscale image meet the distribution of the user's attention.Our image decolorization algorithm consists of four main steps.Firstly,we process the input image to obtain the guidance image with the GrabCut image segmentation technology or the frequency-tuned saliency detection technology.Secondly,we combine the salience of the guidance image and the RGB channels of the input image to build a polynomial parametric model.The salience channel is the key to enhance the contrast locally.Thirdly,construct the energy function,replacing each pixel's output value in the energy function with the parametric model.Next,zoom in or zoom out the standard contrast in the energy function with the guidance image.Finally,by transforming the energy function into partial differential equations,we get the output weight coefficient of each term in the parametric model.When compared with the results of other decolorization methods,our results demonstrate that our algorithm can enhance the contrast of local region well.It can make people understand the context of the grayscale picture more easily and it improves the decolorization quality.