Colorization of Grayscale Images Based on Automatic Content Association

In view of solving the color distortion and nonconformity with subjective feeling in gray image coloring, a new coloring algorithm for gray image based on automatic content correlation is proposed in this paper. First, the feature points of the color reference image is extracted and extended to the 3×3 neighborhood region. Then, a feature set consistent with integrated visual saliency is constructed. Subsequently, the source gray image is matched by structural similarity index measurement (SSIM) one by one in the way of line scanning. And the Cb and Cr of the feature points with the highest matching degree is transferred to the gray image. Finally, the color image is reconstructed in the YCbCr color space and then converted to the RGB color space. The method proposed in this paper does not need user interaction. Compared with other coloring algorithms, it is verified by experiments that our algorithm can effectively improve the visual effect of gray image coloring.