An Approach for Automatic Colorization of Grayscale Images

In this paper a novel approach for colorization grayscale images is proposed. The proposed technique colorizes grayscale images using a reference color image that has a similar mood. A feed forward artificial neural network (ANN) is used. The proposed technique contains two stages, first one is training stage where the system use the reference image to train the system. The output of this stage is three trained neural networks. The second stage is the coloring stage where the system uses the three trained neural networks to color the input grayscale image. Peek Signal Noise Ratio - PSNR quality measure is used to evaluate the proposed technique, and the results showed that typical values for the PSNR which it is between 30 and 50 dB is achieved for recolored grayscale images using proposed technique