Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!seismo!rutgers!ames!ucbcad!ucbvax!ucsfcgl!pixar!ph
From: ph@pixar.UUCP (Paul Heckbert)
Newsgroups: comp.graphics
Subject: Re: A few questions
Message-ID: <476@pixar.UUCP>
Date: Tue, 20-Jan-87 02:34:29 EST
Article-I.D.: pixar.476
Posted: Tue Jan 20 02:34:29 1987
Date-Received: Tue, 20-Jan-87 23:43:19 EST
References: <563@vu-vlsi.UUCP>
Organization: Pixar -- Marin County, California
Lines: 25
Keywords: color distribution
Summary: compression vs. quantization

In article <563@vu-vlsi.UUCP>, williams@vu-vlsi.UUCP (Thomas Williams) asks
how to quantize a 24-bit image down to 8 bits.  And Ed Falk is right
regarding the popularity of this topic: it does come up extremely often.
I count five independent queries about it here in the past 2 years!

Skip Montanaro's statement that DeFanti's technique can present a 24-bit
image on a 6-bit frame buffer isn't quite correct.  What I did in my SIGGRAPH
'82 paper was quantization for display on frame buffers with a colormap,
but DeFanti's SIGGRAPH '86 paper is about image compression.
My goal was display on a limited frame buffer, while his was compact storage
and transmission.

For each 4x4 block of pixels DeFanti transmits a bitmap (16 bits) plus two 8-bit
pixel values which index into a colormap, so that's 32 bits / 16 pixels =
2 bits per pixel, plus one colormap per picture.  It takes some hardware or
software processing to convert his coded blocks into pixel values.
His technique is more akin to other compression techniques (Huffman coding,
transform coding, etc.) than it is to simple quantization.

'Til next time,

Paul Heckbert
Pixar				415-258-2303
P.O. Box 13719			UUCP: {sun,ucbvax}!pixar!ph
San Rafael, CA 94913		ARPA: ph%pixar.uucp@ucbvax.berkeley.edu