COLOR-IMAGE QUANTIZATION WITH USE OF A FAST BINARY SPLITTING TECHNIQUE

We investigate an efficient color-image quantization technique that is based on an existing binary splitting algorithm [IEEE Trans. Signal Process. 39, 2677 (1991)]. The algorithm sequentially splits the color space into polytopal regions and picks a palette color from each region. As originally proposed, the complexity of this algorithm is a function of the image size. We introduce a fast histogramming step so that the complexity will depend only on the number of distinct image colors. Data structures are employed that permit the storage of a full-color histogram at moderate memory cost. In addition, we apply a prequantization step that reduces the number of initial image colors while preserving image quality along visually important color coordinates. Finally, we incorporate a spatial-activity measure to reflect the increased sensitivity of the human observer to quantization errors in smooth image regions. This technique preserves the quantitative and qualitative performance of the original binary splitting algorithm while considerably reducing the computation time.