JPG Decoding

[Dogway]

I was just doing some tests and to my surprise jpg decoding in avisynth wasn't the same as photoshop's, it could be asked the other way around, but since Im on doom9 I ask it from this point of view.
I run some tests and I got some results. Jpg decoding in avisynth using Imagesource, and immaread were the same as default windows Image Viewer. And different to photoshop's as much as (look difference mask below):

http://i212.photobucket.com/albums/cc35/Dogway/Misc/JPG-DecodingDifference.png

My question is just for asking if imagesource is doing the right thing or not

[Didée]

Can't tell you that. But it reminds me of a situation in my company, happened just yesterday.

We got six JPGs from a customer, which had to be included/used by a 3rd company software module (for label printing). Three of the JPGs worked, the other three didn't show a picture in the label software. I looked into the EXIF data, and found that the three non-working pictures were created by Adobe PhotoShop (OS: Macintosh.)
Converted the $$$$ Application pictures with a 0$ Freeware (IrfanView), and the problem was solved. :D

[Dogway]

I come from the CG field where softwares like photoshop or painter are necessary. Im not sold on any kind of software, Im just trying to decide for what things should use one or another.

[pandy]

My question is: is Photoshop doing JPEG decoding correctly? - use DJPEG - reference jpeg decoder and compare results - which one is more close to the DJPEG?

[Dogway]

Yes, that was the other point of view of my question, quite expectable since it was the only one with different results. I just used DJPEG and is the same as imagesource. Thanks for the help! : D

[Dogway]

Seems that Photoshop uses old IJL (Intel Jpeg Library), while most programs use IJG (Independent jpeg group) library. IJL now is dropped and instead there's Intel IPP (Integrated Performance Primitives). Does anyone know where can I test any implementation of this last library? I always heard in doom9 that decoding cannot be good or bad, but instead decodes, or no decodes. But asking in another forum I was told that "JPEG has a lot of leeway in how the images are reconstructed".
btw I don't mind if the thread is moved to General Discussion, since it has got another course.

[amtm]

I always heard in doom9 that decoding cannot be good or bad, but instead decodes, or no decodes. But asking in another forum I was told that "JPEG has a lot of leeway in how the images are reconstructed".

Are you sure you aren't just wrongfully conflating what is said about H.264 decoding to JPEG decoding? Can you link to where someone made that claim with respect to JPEG? The second statement, though, is the correct one. JPEG decoding in the standards has precision requirements but does not have requirements that all decoders output identically. This is the same as the case for mpeg-2 or mpeg-4 asp video formats, for example.