Palette-Based Recoloring of Natural Images Under Different Illumination

Palette-based image recoloring is a popular method in the field of color processing by Intuitively converting the main color of the image into a palette that the user can directly manipulate. However, the traditional palette-based image recoloring method for natural images under different illuminations is affected by the light, colors on the extracted palette may be too close, which will cause color overflow and serious distortion after recoloring. This paper proposes a palette-based recoloring algorithm based on intrinsic decomposition. By using an intrinsic decomposition system based on different illumination images for unsupervised deep learning, natural image is decomposed into reflectance image and shading image. Only the reflectance image is recolored, and then combined with the shading image to obtain a natural recolored image. At the same time, in order to balance accuracy and ease of operation, we use the gap statistic algorithm to get the palette size. The recoloring results of natural images show the superiority of this method in terms of color transfer quality and user experience.