International Journal of Computer Applications (0975 – 8887) Volume 116 – No. 15, April 2015 29 Smooth Context based Color Transfer Dao Nam Anh Department of Information Technology Electric Power University 235 Hoang Quoc Viet road Hanoi, Vietnam a. Input image b. Smooth context by bilateral filter on top,and target color at the bottom c. Result, SSIM=0.9998, t=4.05 Figure 1:Color transfer in smooth context ABSTRACT Color transfer is an emerging framework for dealing with ubiquitous color manipulation in media such as documents and images. Despite the notable progress made in the field, there remains a need for designers that can represent the same information in personalization and corresponding to media context. This work presents adaptive color transfer method using cross-disciplinary interaction of semantic context and bilateral filters. Colors in the method are transferred softly in matching with saliency distributed context. Preliminary results show that the framework is highly keeping consistency and promising. Consequently in this work, a solution of tone mapping by color transfer is introduced. Experimental results are further showed pertaining for automatic handling colors and contrast. General Terms Pattern Recognition, Algorithms Keywords Context, smooth, color transfer, bilateral filter, saliency, tone mapping 1. INTRODUCTION As digital camera technology has advanced in the past decade, colors manipulation, and color transfer in particular is still fundamental and strong basics for several image analysis applications. The term color transfer is used here specifically to the transferring the color style from the target image to the source image. Reinhard et al in [1] provides a comprehensive overview of simple solution to impose one image’s color characteristics on another, where color correction is achieved by choosing an appropriate source image and applying its characteristic to another image. The use of statistical analysis can greatly simplify the computational load of learning colors, provided it can be successfully matched and optimized for the problem at hand. Recent works by [2], [3], [4], [5] have made great strides in scaling both automatic and custom-handled color transfer, though context consistency of these models remains unremarkable. In addition, recent saliency findings suggest that salient regions should contain not only the prominent objects but also the parts of the background that convey the context [6]. Such salient regions facilitate effective matching and interpretation of visual information, particularly in the context of capturing spatio-color dependencies. In fact, if a stimulus is insufficiently salient, sky and water in a scene as in Fig.1a are not changed in color transfer result as in Fig.1c. In this case, saliency map displayed in upper part ofFig.1b is fairly reasonable. Target color as in bottom part asofFig.1b is applied mainly to foreground by the saliency map. The map is smooth but it keeps major features by bilateral filter [9]. The result looks more natural with new color transferred into foreground while keeping background mostly the same. Structural similarity (SSIM) index [7] is 0.9998 for the case. This work has several technical contributions. Firstly, new algorithm of Smooth Context Based Color Transfer (SCCT) is proposed and analyzed. This is a development of statistics- based method [8] for color distribution transfer. The method used different color distribution to find mapping relations for color transfer. This method now is combined with context- aware saliency [6] and bilateral filter [9] in our work to produce a new way of color transfer. Secondly, a solution for tone mapping is suggested. This is the practical implementation the SCCT algorithm for high dynamic range images. Though the solution is not fully automatic but it’s simple, effective and can produce creative results in artistic styles following artist’s reference. 2. OUTLINE OF PAPER In the following sections core concepts are reviewed pertaining to the smooth context based color transfer. The main color transfer mechanisms are discussed along with key metrics and parameters for formulating those mechanisms. Then the experiments are discussed for the McGill Calibrated Color Image Database [10] and results of using SCCT as a solution for the color transfer are demonstrated. High dynamic range images [11] then are tested with tone mapping solution by SCCT. Finally, the article is concluded with discussion and summary of projected future directions for our work. 3. PRIOR WORK On the part to follow the article goals and approach it is helpful to review shortly some major color transfer methods. The task of color transfer still get a lot of attention since handling colors is essential in image analysis. Therefore, in