Message-Id: <mailto:199503132150.PAA20954@library.wustl.edu> Date: Mon, 13 Mar 1995 10:44:39 -0500 From: "mailto:Walter_Gilbert@UMAIL.UMD.EDU--Maryland" <WALT@UMDD.bitnet> Subject: Re: JPEG and image size To: Multiple recipients of list IMAGELIB
On Fri, 10 Mar 1995 14:05:34 -0500 <mailto:KIM@RADSL.MHS.HARVARD.EDU> said: >Hello.
> I'm afraid I don't quite know how to phrase this question, but here
>goes. I know that when you use JPEG you can choose what degree to which
>the algorithm will compress a file, and this degree affects file size and
>image quality in opposite directions. 1. What degree, or percentage, or
>whatever the proper term is, have people used without losing an
>unacceptable amount of image information, 2. can someone give me an
>example of the change in file size using this degree.
>
Kim, A good starting point for JPEG is a "Q" of 75. However, since JPEG is a complex encoding process that takes advantage of certain insensitivities of your eye, the results are dependent on the actual image. Therefore, you may find that for some images Qu is OK, for others you may need Q…, others may be OK with Q`. My experience shows about a factor of 10 to 20 reduction in image size at Qu; that is, a 1MB image is reduced to 100K to 50K. At Q0 (lossless), expect about a 3:1 to 4:1 compression: 1MB reduces to 250K-350K. However, note that JPEG is NOT good for line drawings (except at Q0 where you don't get very good compression). Since JPEG works on 8x8 pixel hunks, if you compress pictures with large areas of slowly varying colors, as a sky, you can sometimes see artifacts of the 8x8 hunks. If this happens, you need to increase the Q value. WaltWalter Gilbert, Asst. Dir. mailto:Walter_Gilbert@umail.umd.edu Computer Science Center Manager: Teaching Technologies University of Maryland at College Park 20742-2411 (301)405-6727