Digital halftoning based on color correction using neural network with uniform color samples and vector error diffusion

This paper proposes a uniform color sample selection and color halftoning method based on color correction using neural network (NN) with a set of uniform color samples and selective vector error diffusion for enhancing color reproduction on a printer. In order to generate uniform color samples in CIELAB color space, a set of uniformly populated color samples in a CIELAB printer gamut and monitor gamut are calculated by LEG (Linde, Buzo, Gray) quantization algorithm. Then, the corresponding device-dependent values of CMY and RGB are estimated by a trained NN, which was temporally trained by a set of uniform samples in the device-dependent spaces. The estimated sample colors in the device-dependent spaces are utilized as inputs to produce their real colorimetric values in terms of device-independent colors, CIELAB values. Device dependent and independent pairs of the generated uniform color samples are exploited to train the second NN. And the trained NN is utilized to estimate output colors of a monitor and printer in the halftoning process. In the halftoning process, the color of each pixel of an image is estimated by the second neural system and the input color is corrected by selective vector error diffusion to minimize a colorimetric color difference between a printer and monitor. Finally, the printed colors exhibit a better color reproduction than the conventional scalar color halftoning and were well matched with the monitor colors without additional gamut mapping technique.