dickey@ssc-vax.UUCP (Frederick J Dickey) (11/16/87)
Does anybody have any pointers to reverse dithering techniques? In
particular, suppose you have a bilevel display device and you want
to display an image that has several gray scales. A common technique
seems to be to "dither" the image with patterns like the following:
. . .....
. ......
. . .....
My question is the following: Given the image in the dithered form,
is there any reasonable way of reconstructing the original gray scale
image?cmcmanis@pepper.UUCP (11/18/87)
In article <1525@ssc-vax.UUCP> dickey@ssc-vax.UUCP (Frederick J Dickey) writes: >My question is the following: Given the image in the dithered form, >is there any reasonable way of reconstructing the original gray scale >image? If you know the region of the dither (like 4 X 4 squares or something) you can just average all of the pixels in the region and multiply that by the a normalizing factor. --Chuck McManis uucp: {anywhere}!sun!cmcmanis BIX: cmcmanis ARPAnet: cmcmanis@sun.com These opinions are my own and no one elses, but you knew that didn't you.
holloway@drivax.UUCP (11/18/87)
In article <1525@ssc-vax.UUCP> dickey@ssc-vax.UUCP (Frederick J Dickey) writes: >My question is the following: Given the image in the dithered form, >is there any reasonable way of reconstructing the original gray scale >image? 1) Find the size of the dithering matrix. This should be pretty obvious, and usually square (8x8 pixels is common) 2) Take each pixel in turn as the center of the dithering matrix. Then count the number of white pixels in the matrix, divide by the total number of pixels in the matrix, and you have your percentage of white for that pixel (from 0 to 1). This method isn't foolproof, especially where the intensity changes quickly, but it should serve. I'd be interested in hearing about any better algorithms. - Bruce -- ******************************************************************************* * Bruce Holloway - Terminal Netnews Addict uunet!amdahl!drivax!holloway * * ALBATROSS, ATARI*TROS @ Plink ALBATROSS @ Delphi * *******************************************************************************
flaig@cit-vlsi.UUCP (11/18/87)
In article <34159@sun.uucp> cmcmanis@sun.UUCP (Chuck McManis) writes: >In article <1525@ssc-vax.UUCP> dickey@ssc-vax.UUCP (Frederick J Dickey) writes: >>My question is the following: Given the image in the dithered form, >>is there any reasonable way of reconstructing the original gray scale >>image? > >If you know the region of the dither (like 4 X 4 squares or something) you >can just average all of the pixels in the region and multiply that by the >a normalizing factor. >--Chuck McManis Unfortunately, while the dithering _pattern_ may be 4x4, most dithering is done on a pixel by pixel basis. Information (in the form of more bits of color resolution) is thrown away when a picture is dithered, and cannot be recovered (maybe that pattern of dots _was_ in the original picture). There are two exceptions where information can be recovered: 1) If the dithering wasn't done on a pixel by pixel basis (usually this is only done with low resolution pixtures, and the result shows it). 2) If you can identify large areas in the dithered picture which were a constant untextured shade in the original, then averaging pixels within that area will give the original color. But most interesting pictures have textures or shading which will not allow this. Note: I am not an expert in the above topic, but I have taken classes in graphics and information theory and done dithering myself. --Charles Flaig flaig@vlsi.caltech.edu
jbm@aurora.UUCP (Jeffrey Mulligan) (11/19/87)
in article <1525@ssc-vax.UUCP>, dickey@ssc-vax.UUCP (Frederick J Dickey) says:
+
+
+ My question is the following: Given the image in the dithered form,
+ is there any reasonable way of reconstructing the original gray scale
+ image?
Do what the eye does: low-pass filter.
--
Jeff Mulligan (jbm@ames-aurora.arpa)
NASA/Ames Research Ctr., Mail Stop 239-3, Moffet Field CA, 94035
(415) 694-5150marinell@dalcs.UUCP (Kevin Marinelli) (11/19/87)
In article <4579@cit-vax.Caltech.Edu> flaig@cit-vlsi.UUCP (Charles M. Flaig) writes: >In article <34159@sun.uucp> cmcmanis@sun.UUCP (Chuck McManis) writes: >>In article <1525@ssc-vax.UUCP> dickey@ssc-vax.UUCP (Frederick J Dickey) writes: >>>My question is the following: Given the image in the dithered form, >>>is there any reasonable way of reconstructing the original gray scale >>>image? >> >>If you know the region of the dither (like 4 X 4 squares or something) you >>can just average all of the pixels in the region and multiply that by the >>a normalizing factor. >>--Chuck McManis I have tried to do something similar myself recently, but by using grey levels only. I have used two techniques, both however have their problems. It is assumed that the dithered image uses 1 bit per pixel, although this may be altered if necessary. Algorithm 1 for each pixel in the image, the grey level = sum of the bits that are on for each of its neighbors, and itself. This gives a 9 level grey map. (this may be extended to neighbors within 2 pixels for a 16 level grey map. This technique preserves the spacial resolution of the dithered image, but causes the grey level map to look "Blurred". In many cases this is not a problem. If there are a lot of high contrast edges, then the bluring is more pronounced. E.g for black text on a white background. Algorithm 2 for every block of N X N bits, the grey level is the sum of the on bits. The next pixel in the sequence is the next adjacent block of N X N bits. This technique compresses the spacial resolution of the dithered image to 1/N. Bluring of the image is much less of a problem. This works best for very large dithered images where the loss of spacial resolution is not important. I have used the above technique for converting images from an apple Mac for display on an IBM PGA card. Because the images on the Mac were monochromatic dithered images, it was not possible to restore any actual colour information to the image. -- Kevin Marinelli Academic Computing Services ! marinell@dal.bitnet BITNET Dalhousie University ! marinell@dalcs.UUCP UUCP Halifax, Nova Scotia, CANADA ! marinell%dal.bitnet@wiscvm.wisc.edu INTERNET