Lossless rotation is not lossless

[Luca]

Hello, i'm novice to XNVIEW and i'm trying to use it for my needs.

I found that lossless rotation is not lossless... i verified this problem using the procedure described at

http://www.impulseadventure.com/photo/l ... -test.html

I've already used this procedure for testing other tools (for example AcdSee passed the test, JPEG LossLess Rotator didn't etc.).

Bye.

Luca

[marsh]

That linked reference is misleading because it prompts visitor to 'resave and compare to original'. Saving jpeg according to quality settings (ex. 75%) should not be confused with a lossless jpeg operation which directly modifies file. It is better to use lossless rotate tool without resaving image.

[XnTriq]

• minor change when lossless saved?
• Lossless JPEG transformations
• 1.91.6: Lossless JPG rotation not lossless?

[Luca]

Sorry, but i didn't save anything (directly).

I wanted to use XnView as a browser, for rotating images that need it. So this is what i did:

- i selected my folder in the left tree so all my pictures where displayed on the top right as thumbnails
- i clicked a picture and it was displayed on bottom right preview area
- i selected Outils/Transformations JPEG sans perte/Rotations 90° à droite (4 times in this case for obtaining a picture with same rotation as the original)
- than i made the comparison test (i previously made a copy of the original picture) describred in the link mentioned in previous e-mail with faulty result

If you make the exact same sequence with AcdSee you will obtain a perfectly equal picture as the original.

You can say: USE ACDSEE (i have a regular copy of it)!

But i think that your tool is better and this is the reason i'm wasting my time writing this e-mail.

Thanks.

Luca

[ckv]

I can reproduce the file difference by rotating the file 4 times, but the result still isn't lossy. After rotaing the file, I saved the original image and the rotated copy as windows bitmap (any lossless format should be fine) and then compared the bitmaps instead of the jpgs. This way I exluded all the differences that that doesn't effect the actual image data (e.g. metadata).