A Fast and Practical CNN Method for Artful Image Regeneration

Although artists’ actions in photo retouching appear to be highly nonlinear in nature and very difficult to characterize analytically, we find that the net effects of interactively editing a mundane image to a desired appearance can be modeled, in most cases, by a parametric monotonically non-decreasing global tone mapping function in the luminance axis and by a global affine transform in the chrominance plane. This allows us to greatly simplify the existing CNN methods for mimicking the artists in photo retouching, and design a new artful image regeneration network (AIRNet). The objective of AIRNet is to learn the image-dependent parameters of the luminance tone mapping function and the affine chrominance transform, rather than learning the end-to-end pixel level mapping as in the standard practice of current CNN methods for image restoration and enhancement. The proposed new approach reduces the complexity of the neural network by two orders of magnitude, and as a side benefit, it also improves the robustness and the generation capability at the inference stage.