Abstract
In this project we have described a method of colorizing a gray scale image based on segmented reference image. This method requires considerably less human intervention. The user here just need to provide a colored image which will serve as a reference to the algorithm. The objects and texture content of the reference image must be as similar as it can be to the target image. Although we can never be perfectly sure that matching of objects from target to reference image is absolutely correct. Hence to improve this we partially colorize the given image at those points where we are absolutely sure.Image colorization is process of recovering color information from given only the luminous intensity information. Mathematically we have to create three dimension RGB data from one dimension intensity data, for each pixels in the target gray image. This information cannot be created on its own, we need to provide some previously learned information
Result from colorization
Reference Image Target Image Resultant Image