DDColor is a photo-realistic image coloring tool that leverages dual-decoder technology to autonomously learn image content, achieving highly realistic coloring effects for historical black-and-white photos and anime scenes. Its working principle involves an encoder analyzing the image, multi-scale feature extraction, coloring decision-making, and the decoder's output integration, while also introducing a color richness loss function to innovatively enhance the color saturation and appeal of the generated images. For more detailed information, please visit the project's website.