Compression of Monochrome and Color Images
The application of the approximation to image compression consists first of
representing the image in terms of regions and contours, then encoding of
these two components. In application which require higher visual quality,
the error between the original image and its approximation is also
encoded.
In order to reduce the number of bits required for encoding the error, it is
quantized using a logarithmic quantizer designed to quantize small errors
more finely than the large one (to exploit the visual masking around the
edges). The encoder block diagram is shown in Figure 5. As a further
refinement, a separate quantizer is used for smoothly varying regions. This
allows us to allocate more bits to error in the smoothly varying regions.
This adaptive quantization scheme results in a substantial compression of
the error data.