wan@mercury (Shijie Wan) (04/22/91)
Color Compression Software Available ------------------------------------ A full-color image can use up to 16 million different colors. Most commonly used color monitors and laser printers can display/print only 256 or less different colors simultaneously. In order to display/print a full-color image on these devices, the number of colors in the original image must be compressed before display. We have designed a color image quantization algorithm which can reduce the colors in a full-color image to 256 colors or less with minimum distortion. Experiments show that in most cases our algorithm needs to use 128-256 colors to produce an image which looks almost identical to the original full-color image. With 32-64 colors, it can produce images of high quality. For 32 colors or less, it can generate quantized images with minimum contouring effects, with the aid of dithering techniques. For details of the algorithm and image examples, read the paper: S.J. Wan, P. Prusinkiewicz, and S.K.M. Wong, Variance-based color image quantization for frame buffer display, Color Research and Application, vol. 15 (1), pp. 52-58, Feb. 1990. The program is written in C for Silicon Graphics IRIS machine. Versions for other machines running UNIX are also available. Excluding the time for image disply, the program needs about 10-40 seconds to process one image. For more information, please contact: ---------------------------------------------------------------- | wan@mercury.uregina.ca | Shijie Wan | | Tel: (306)-584-5717 | Dept. of Comp. Sci. | | (306)-585-4690 | Univ. of Regina | | Fax: (306)-584-5717 | Regina, Sask. | | | Canada S4S 0A2 | ----------------------------------------------------------------
craig@weedeater.math.yale.edu (Craig Kolb) (05/03/91)
In article <52@mport.COM> lee@mport.COM (Lee Crocker) writes: >wan@mercury (Shijie Wan) writes: > >> Color Compression Software Available > >> [Description of an implementation of the WWP vector quantizer.] > >There are several public domain sources for this algorithm, most of them >based on an implementation by Craig Kolb. The FBM utilities is probably >the best source for Unix. Piclab implements the algorithm for MSDOS. >Both are available FTP from various graphics-oriented archive sites. I should set the record straight by saying that, as far as I know, the Utah Raster Toolkit is the only package that incorporates my code. The credit for "fbquant" and whatever is implemented in piclab goes to their respective authors. The code I wrote, "colorquant", implements the algorithm described in: Wan, Wong, and Prusinkiewicz, An Algorithm for Multidimensional Data Clustering, Transactions on Mathematical Software, Vol. 14 #2 (June, 1988), pp. 153-162. Colorquant is available via anonymous ftp from weedeater.math.yale.edu as pub/colorquant.shar. The Utah Raster Toolkit is available from weedeater.math.yale.edu, freebie.engin.umich.edu, cs.utah.edu, and other archives sites. Cheers, Craig
lee@mport.COM (Lee Crocker) (05/04/91)
Having examined your code from the Utah Raster Toolkit and that from FBM, it is clear that the latter was based on your design. Having written the code in Piclab, I can state definitively that I based it on your design as well as the ACM paper. I did not use your code directly (indeed I could not as it required much more memory than MSDOS can handle), but I was definitely "inspired" by it. Piclab is available from cica.cica.indiana.edu and wsmr-simtel20.army.mil. -- Lee Daniel Crocker | "Computers are useless. They can only give lee@mport.com (Microport) | answers." --Pablo Picasso ...!uunet!mport!lee 73407.2030@compuserve.com