A texture recognition coloring technique for natural gray images

Gray image coloring is utilized to increase the visual appeal of images such as old black and white photos, movies or scientific illustrations. Most of authors working in coloring of gray scale images have used primitive methods for coloring which are both inaccurate and limited. In this paper we propose a new technique for computer coloring gray scale images. This technique works for texture based images like natural scenes. It's based on segmenting the image into different regions according to their textures. Real colors of the image can then be obtained by classifying these textures to predefined texture classes. Recognition of these textures is performed by matching these textures with a training set stored in a special database. We validate the efficiency of our coloring system by coloring several sets of natural gray images with high quality real colors.