Global Distribution Adjustment and Nonlinear Feature Transformation for Automatic Colorization

Figure 8

(a) The outputs colorized images using our proposed method for animals (C), Ground Truth (A), the gray image of which is an input image, and reference image (B)

(b) The outputs colorized images using our proposed method for other creations (bird, insect, and crocodile) (C), Ground Truth (A), the gray image of which is an input image, and reference image (B)

(c) The outputs colorized images using our proposed method for artificial objects (helicopter, chair, sailboat, and folk art object) (C), Ground Truth (A), the gray image of which is an input image, and reference image (B)

(d) The outputs colorized images using our proposed method for humans (C), Ground Truth (A), the gray image of which is an input image, and reference image (B). We find that human’s clothes sometimes fail to get correct colors