A Fast Color Synthesis Algorithm Using the l-α-β Color Space and a Non-parametric MRF Model

We present a fast method for adding color to grayscale images. The method uses a statistical learning approach based on changes in intensity to color by example. Typically, these approaches take a considerable amount of time, proportional to the image size. To reduce the computational time, we propose a framework that first obtains a synthesized color image of a scaled version of the input grayscale image. Then, it scales up the synthesized color together with the high-frequencies (luminance) of the image original size by using the l-α-β color space. Preliminary results on a set of nature images are shown.