Quantization of color image using generic roughness measure

Color quantization is a technique which is used to compress the color space of an image to reduce the visual distortion. The computational complexity of preclustering based quantization is less, but not guaranteed the quantization precision. The quantization quality is high in post clustering based quantization but computational complexity is high. In color quantization the balancing of quantization quality and complexity is very challenging thing. To compensate this, two stage quantization framework is proposed. In first stage, the color space with high resolution is compressed into a color space with condensed type by thresholding. For that, we propose generic roughness measure for effective segmentation of color image. In second stage, the compression results are clustered to form a palette by k-means clustering to get the final results.