Message-Id: <mailto:199408292337.SAA04437@library.wustl.edu> Date: Mon, 29 Aug 1994 16:36:48 EDT From: Robert Rosenberg <mailto:rarosenb@GANDALF.RUTGERS.EDU> Subject: Re: Image database and Huffman Encoding To: Multiple recipients of list IMAGELIB <mailto:IMAGELIB@ARIZVM1.BITNET>
There is a very neat and clear description of Group 3 and Group 4 compression in the August 94 issue (the latest one) of _Imaging_ magazine. Essentially, it is this:Group 3 sends a reference line every four or eight lines. Each reference line is "bbwwwwbbwbwbwwbbbwbwb" where b˙ack and w˙ite (Groups 3 & 4 are only for black and white, or 1-bit, images). For the intervening lines only the changes are sent; for example, if the line after a reference line were "2,5,15,17" then those pixels would switch to black or white, whichever they weren't before. So the received picture is built up one line at a time. If a line gets screwed up in transmission, only four or eight lines at most can be garbled, because then another reference line is sent.
Group 4 sends only one reference line--the first one. So if you don't have an error-correcting line, fax, or faxmodem, one error will garble the rest of your transmission. But the compression is much better, since only changes are sent for the whole document.
(the reference line isn't actually "b" and "w"--it's just representative bits).
Bob Rosenberg | If the Lord had more Thomas A. Edison Papers | respect for money he'd Rutgers University | have given it to New Brunswick, NJ 08903 | a better class of people. mailto:rarosenb@gandalf.rutgers.edu | --Maverick's Pappy