Can I rotate lossily compressed photos that I view in Windows Photo Viewer without worrying about losing even more information to compression?
-
1\$\begingroup\$ Note that--as far as I know--a 90 degree rotation (and 180, 270) is lossless in and of itself because it is a swap of x and y coordinates (other angles require interpolation). So any lossyness attributed to the transform is from the recompression. You can avoid this recomputation by saving the transformed image in a lossless format, but that is clearly not part of your question, so I won't suggest it \$\endgroup\$– horatioCommented May 24, 2011 at 18:32
-
6\$\begingroup\$ @horatio, the multiples of 90 degrees are a special case for JPG compression, where it is technically possible to rotate the image without a compression cycle by "simply" rearranging the compressed data. It does involve removing and applying a layer of lossless compression, so the file size might change, but no lossy operations are required. \$\endgroup\$– RBerteigCommented May 24, 2011 at 18:54
-
1\$\begingroup\$ I think I understand you, but my point is that jpeg is a disk storage format and does not have any bearing on the quality of the transforms themselves. That is to say, though one often speaks of working with jpegs etc, while the file is displayed on screen and manipulated, it is not a jpeg. \$\endgroup\$– horatioCommented May 24, 2011 at 19:59
-
1\$\begingroup\$ Man I have ALWAYS wanted to know the answer to this question. IMO they SHOULD be. But most likely due to the JPEG standard... I don't think it is technically possible to make this operation lossless. \$\endgroup\$– Trevor Boyd SmithCommented May 24, 2011 at 20:46
-
2\$\begingroup\$ Google Picasa image viewer IS lossless. but it cheats. it doesn't rotate the image. it just marks down in meta data "display this image rotated". \$\endgroup\$– Trevor Boyd SmithCommented May 24, 2011 at 20:46
1 Answer
If the dimensions of the image are multiples of 8 (or 16 if chroma subsampling is used) then the rotations are lossless. Otherwise it is not possible to rotate the image without recomputing the blocks i.e. recompressing the image, which is lossy.
The reason for this is that jpeg images are broken up into a series of 8x8 or 16x16 blocks which are compressed individually. Incomplete blocks are only allowed on the right edge and bottom edge. Thus is your image is not an exact multiple of 8/16 it will contain incomplete blocks, which will end up on the wrong edge after rotation.
I verified the above assertion using the version of Windows Photo Viewer that ships with Windows 7. I used two images of colour noise. One image was 256x256 i.e. both sizes multiples of 8. The other image was the same but cropped to 253x253 i.e. neither size was a multiple of 8. Here are the images:
I then performed four anti-clockwise rotations, closing photo viewer after each rotation to ensure the image was saved in it's rotated state.
There's no need to perform a subtraction to see the difference with the 253 pixel image, it's noticeably darker and muddier!
A diff on the other image looks like this:
i.e. the images are identical, the rotations were lossless.
-
1\$\begingroup\$ I know that in the past when I've done rotations it has warned me that the rotation would be lossy, which implies that in cases where the warning wasn't displayed the transformation was lossless, but I didn't check. Nor do I remember what version of WPV I was using, not precluding the possibility that MS changed the codebase, so no I can't be sure. \$\endgroup\$ Commented May 24, 2011 at 13:55
-
1\$\begingroup\$ @Tim an image of pure noise is by definition the worst case for jpeg (or any compression algorithm), which is why I chose it! Results on real images would look better. \$\endgroup\$ Commented May 26, 2011 at 15:12
-
1\$\begingroup\$ @MattGrum, very cool demo. I hope all digital camera makers have their image sizes in integer multiples of 8! \$\endgroup\$ Commented May 27, 2011 at 16:36
-
1\$\begingroup\$ @MattGrum I used to work on the WPV team. AFAIK, (and I'm pretty sure) your answer is correct. Also, +1 for a very thorough method of proving your hypothesis : ) \$\endgroup\$– TristanCommented Aug 6, 2013 at 16:53
-
\$\begingroup\$ @TrevorBoydSmith every camera I've seen has image dimensions that are a multiple of 8 or 16. Rotation isn't necessarily the reason, it may just allow some optimization of the JPEG pipeline. \$\endgroup\$ Commented Oct 10, 2019 at 22:47