Image Quantization effects
Rose in 256 colors
This image compares rose512.jpg saved as 256 color xpm by libAfterImage,
to the same image quantized using XNview. libAfterImage provides the fastest functionality to convert images into
indexed format. Colors are reduced not by closeness as usually done (libXpm) but by allocating colorcells to most used
colors first, and then approximating other colors with those allocated. It is possible to reduce colordepth of the
image prior to quantization, using error diffusion technique. That allows for fewer colors in resulting image without
quality loss, as demonstrated below.
Rose in 256 color reduced to 12bpp
As can be seen here reducing colordepth of an image prior to quantization
actually improves quality of the resulting image. Again image produced with libAfterImage
is on the left.
Rose in 256 color reduced to 9bpp
Even as we reduce colordepth to such an extent as 9bpp we still get rather decent
quality, comparable to conventional dithering.
Rose in 256 color reduced to 6bpp
And only at point of 6bpp, when most of the color info has been striped out we get
sugnificantly worse image. Althou it should be noted that result depends on the image itself,
and for some images even 6bpp should work just fine.