Abstract
The process of converting gray images or videos to color ones by adding colors to them and transforming them from one-dimension to three-dimension is called colorization. This process is often used to make the image appear more visually appealing. The main problem with the colorization process is the lack of knowledge of the true colors of the objects in the picture when it is captured. For that, there is no a unique solution. In the current work, the colorization of gray images is proposed based on the utilization of the YCbCr color space. Reference image (color image) is selected for transferring the color to a gray image. Both color and gray images are transferred to YCbCr color space. Then, the Y value of the gray image is combined with the Cb and Cr values of the reference image, based on the Euclidian distance between them. The quality of the resulted image was measured based on several quality measures, which indicated very good results. The proposed algorithm is simple, efficient, and fast.
Highlights
Image colorization is a major challenge to researchers, since it is a difficult and ill- posed problem with no unique solution
The proposed algorithm was implemented with many gray scale images and the results were obtained using several quality measurements (RMSE, Peak Signal-to-Noise Ratio (PSNR), Average Difference (AD), Maximum Difference (MD), Normalized Absolute Error (NAE), Normalized Cross Correlation (NC), Structural Content (SC), Structure Similarity Index (SSIM), and CC)
For accurate measurement of colorization performance, we suggested to select a gray image that is converted from a known color image, which aids to visually compare the colorized image with the ground truth image, by determining these measurements
Summary
Image colorization is a major challenge to researchers, since it is a difficult and ill- posed problem with no unique solution. There are many methods used for matching the pixels between reference and gray images [2and many researchers suggested colorizing algorithms. The suggested method is called "the example-based method" which transfers the colors from the color image to the gray one. The author attempted to increase the accuracy of transferring the color information to the gray image This method is limited to static objects and scenes and unable to colorize dynamic objects. Chia et al [10] introduced a framework to select the proper reference images from the set of internet images This method needs the user to suggest a semantic text label, which is used to search for proper images in the internet. It costs the user efforts to segment the foreground objects from the background scene
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
