Re: Compressive images test

Hi Yoav,

Basically that's it, many images on the Web don't fit to MCUs because people are no longer familiar with JPEG inner workings or simply don't care about the percent or two of the file size they could save, in the other hand digital camera makers still closely follow this rule (that's perhaps because their engineers don't have the time to leave the fast track and fiddle with data padding).

Nevertheless knowing how JPEG works may help make better educated guesses, who knows what's the difference between quality 50 and 51 in Photoshop Save for Web? who knows that the quality level of luma and chroma can be set totally independently?

Regards
-- 
Frédéric Kayser


Yoav Weiss wrote :

> Hi Frédéric,
> 
> If I'd try to "translate" what you said so it'd be better understood by a wider audience, it would something like:
> 'JPEG images that can be perfectly divided into 8x8 pixel "boxes" will compress better'.
> This is great, and need to backed up by a larger scale research (which is on my todo list :) ), but is also slightly tangent to the issue discussed here.
> 
> Correct me if I'm wrong, but I think the point you tried to make is "The examples we're using are the wrong ones".
> If so, I'm not sure I agree. Since many of the images on the web don't perfectly fit MCUs (The 8x8 pixel "boxes"), I think we should test whatever we want to test on a large number of images, extracted from the Web. Some will be a perfect fit, but some won't.
> 
> Let me know if I got it wrong,
> Yoav

Received on Monday, 16 September 2013 13:31:08 UTC