Research On Two Improved Decolorization Algorithms

In our daily life, gray image has great application value in digital printing, medical research field and so on. Therefore, color to gray transformation is becoming a hotspot in the field of image processing technology research. In theory, color to gray transformation is a process of dimensionality reduction, so losing information is inevitable. As a consequence, the goal of decolorization is to use the limited gray level range to preserve as much information of the original color image as possible.Current decolorization algorithms have two problems in their methods:1) They can not preserve structure and local contrast information commendably.2) They can not preserve contrast, color consistency and grayscale pixel characteristic simultaneously.In order to solve the problem that other methods can not preserve structure and local contrast information commendably, we propose a new decolorization method which is called the optimization algorithm of color-to-gray in gradient domain. In our algorithm, we first linear combine the R, G and B three channels (or directly get the luminance channel) to transform the original image from color to gray. Then we compute the color contrast of original image and combine it with the initial decolorization image to build error energy function. At last, we use iterative method to solve the energy function and obtain the final gray image. The experiment indicates that our method can preserve the global structure and local contrast information better.In order to solving the problem that current methods can not preserve contrast, color consistency and grayscale pixel characteristic simultaneously, we propose a new algorithm, decolorization with contrast, color consistency and grayscale pixel preservation, which can maximally maintain these features of original color image. For preserving the structure and fine detail, we use a bimodal Gaussian distribution, which is followed by the difference between pixel and its neighbors, to construct the error energy function. For global color consistency, we use the locally linear embedding to build energy function which makes the same color pixels have the same gray levels in the result. For grayscale pixel characteristic preservation, we mark out grayscale pixels and specify that the gray values of grayscale pixels are known quantities and unchanged during conversion firstly. Then we construct the energy function between grayscale pixels and other pixels. After that, we build the objective function by the linear combination of the three energy functions and obtain the gray image via solving the objective function with the iterative method. The experiment result shows that our algorithm can not only preserve contrast information but also preserve color consistency characteristic and grayscale pixel characteristic better.