Reducing false contours in quantized color images

The ultimate goal of color image quantization is to minimize visible distortion. We present a feedback-based quantization strategy along with its latest implernentation. The system examines and compares the original and the quantized image to detect visible distortions in order to guide the redistribution of resources, i.e. quantized colors, during re-quantization of the original. Resources are reallocated to areas where the most offensive detected distortions occur. Our current focus is to control the false contouring effect—a common eye-catching artifact that often appears after quantization when the original is a computer-synthesized or high-quality photographic image with gradually-shaded areas. The key difference between our new system and our previous implementation is a quantitative feedback-control mechanism that takes into account the magnitude of a false contour as well as the size and color deviation of the image areas alongside the false contour. The effectiveness of our system is demonstrated using raytraced and photo CD images.