Get it right vs fix it in post.. (Page 22): And I didn't say that did I. You are deceitful. Which is why I said, "...as much as half and often more than half..." The amount of data...

Ysarex Loc: St. Louis

BigDaddy wrote:

Your post, (the part that didn't quote) is misleading. 50% of data is not changed,

And I didn't say that did I. You are deceitful.

BigDaddy wrote:

the amount changed is dependent on compression level selected and the make up of the image.

Which is why I said, "...as much as half and often more than half..."

BigDaddy wrote:

Most of the time at low compression levels it's completely insignificant, and is why jpg is used everywhere.

The amount of data changed even at low compression rates is not insignificant. I just took an 8 bit TIFF file and saved it as a JPEG at the best possible (PS) JPEG quality. The TIFF file was 34.3 megabytes and the resulting JPEG is 4.7 megabytes (see below). An 86% compression rate is not insignificant and required a significant amount of data alteration -- you have no idea what you are talking about.

BigDaddy wrote:

You are correct that it becomes a question of standards or tolerance and how the image is used. That's exactly why shooting raw all the time is overkill for most photography done today.

For you. You have not been elected a spokesperson for everyone.

BigDaddy wrote:

Blindly telling people you can't successfully edit jpg images is frankly stupid.

I would never tell people that. Why would you say dumb stuff like that?

BigDaddy wrote:

The world is packed with fantastic images shot as jpgs, and edited as jpgs. And whilst I'm at it, shot in auto white balance with no one but you claiming to notice...

If you search the topic you'll find many people notice. Your exaggerations align with your incompetence.

(Download)

dsmeltz Loc: Philadelphia

JohnSwanda wrote:

For me, bracketing is only useful for non-moving subjects. If you use bracketing to shoot people, for example, what are the chances that the photo with the best expression will also be the one with the best exposure?

Depends on how good you are in post.
Have you heard of "layers"?? Or are you just an "adjust a little" in post type?

selmslie Loc: Fernandina Beach, FL, USA

Ysarex wrote:

Tone/color alterations can cause degradation when an 8 bit image is edited. Tone/color alterations also interact with the JPEG compression grid which produces a different type of degradation. That's what I originally noted.

There are all sorts of problems that can be introduced when the JPEG is compressed again after editing a previously compressed JPEG.

There are several edits we haven't mentioned yet that can lead to problems with the second compression. They are described in the concluding remarks (pages 19 and 20) of Aligned and Non-Aligned Double JPEG Detection Using Convolutional Neural Networks. The rest of the article is not for the faint of heart.

It helps if the grid before and after editing is aligned and, if not, the misalignment can be detected and accounted for. If that can't be done the problems will be worse.

Ysarex Loc: St. Louis

selmslie wrote:

There are all sorts of problems that can be introduced when the JPEG is compressed again after editing a previously compressed JPEG.

I wasn't referring to problems that result from recompression and resaving the JPEG. Those are different problems.

JohnSwanda Loc: San Francisco

dsmeltz wrote:

Depends on how good you are in post.
Have you heard of "layers"?? Or are you just an "adjust a little" in post type?

I do a lot of post processing, mostly layer based in Photoshop. I don't know what that has to do with determining the correct exposure to shoot moving subjects rather than bracketing.

selmslie Loc: Fernandina Beach, FL, USA

Ysarex wrote:

I wasn't referring to problems that result from recompression and resaving the JPEG. Those are different problems.

So how do you edit a JPEG and introduce problems if you don't save it after editing? By saving it as a TIFF?

DelRae Loc: Oregon

aggeorge wrote:

Paul,

I just want to say that your responses to posts are always respectful and framed with the goal of increasing knowledge. Every time I read one of your responses I am just a little bit (sometimes a bunch) better amature
photographer than I was before. Thank you!!!

Alan

I agree with aggeorge 100% and I also THANK YOU DelRae

Ysarex Loc: St. Louis

selmslie wrote:

So how do you edit a JPEG and introduce problems if you don't save it after editing?

By making changes to the tone/color of the image -- the editing creates problems as the editing changes interact with the JPEG compression grid that's embedded in the image. Save it as anything you want, the editing creates JPEG unique degradation.

Urnst Loc: Brownsville, Texas

selmslie Loc: Fernandina Beach, FL, USA

Ysarex wrote:

By making changes to the tone/color of the image -- the editing creates problems as the editing changes interact with the JPEG compression grid that's embedded in the image. Save it as anything you want, the editing creates JPEG unique degradation.

If you load it and convert it to a TIFF (8- or 16-bit) the JPEG compression grid loses its impact on the image. It no longer serves any purpose.

If you then edit the TIFF (preferably the 16-bit version) the compression grid, if it still exists, is not used. If you then decide to save it as a JPEG you are going to get a new compression grid just as if the session had originated from raw or a camera's TIFF SOOC.

JD750 Loc: SoCal

Ysarex wrote:

By making changes to the tone/color of the image -- the editing creates problems as the editing changes interact with the JPEG compression grid that's embedded in the image. Save it as anything you want, the editing creates JPEG unique degradation.

?

JPEG noise increases with every save.
But of course it’s still possible to make limited changes to a JPEG and if the changes are saved only once the noise effect is not noticeable.

What is this JPEG specific degradation you refer to? Can you post an example and describe the conditions that caused the problem?

Ysarex Loc: St. Louis

JD750 wrote:

?
What is this JPEG specific degradation you refer to? Can you post an example and describe the conditions that caused the problem?

JPEG works by laying an 8x8 pixel grid over the image. Each grid cell then contains 64 pixels. Photo data is dense and odds are those are 64 unique pixels. JPEG achieves it's compression rate by very effectively altering pixels in each grid cell to create redundancy. When it's finished a pixel cell will for example go from 64 unique pixels to just 24 unique pixels. JPEG does a great job of this and a great job of blending all the grid cells together so that when we look at the image we see it as unchanged. JPEG as such is terrific technology.

But those altered grid cells are now part of the image. If you go back and start changing tone/color in the image to a significant degree those tone/color changes will interact in rather unfortunate ways with the compression grid that's embedded in the image. Basically the changes start to make the grid visible -- it lifts up out of the image as it were.

This is less of problem in recent years as we've increased camera resolution and effectively swamped the degradation under increased resolution so we don't see it now with 24 megapixel images like we did with 8 megapixel images.

I'd be happy to show you an example but you'll have to wait till this afternoon as I'm off to run some errands.

Ysarex Loc: St. Louis

selmslie wrote:

If you load it and convert it to a TIFF (8- or 16-bit) the JPEG compression grid loses its impact on the image.

It does not. The compression grid that was originally used to create that JPEG is still there embedded in the image. That's the compression grid that interacts with editing changes. Converting to a TIFF in no way alters that exiting embedded compression grid.

JD750 Loc: SoCal

Ysarex wrote:

JPEG works by laying an 8x8 pixel grid over the im... (show quote)

Great! Thanks.

selmslie Loc: Fernandina Beach, FL, USA

Ysarex wrote:

It does not. The compression grid that was originally used to create that JPEG is still there embedded in the image. That's the compression grid that interacts with editing changes. Converting to a TIFF in no way alters that exiting embedded compression grid.

I didn't say that the conversion to a TIFF alters the compression grid. However, it renders it obsolete and useless if you continue to edit the TIFF. And it is not used when you save the image as a TIFF or any other non-JPEG format (PSD, BMP, GIF, PSE, PNG, etc.).

So there is no point in having the editor on your computer wasting cycles on keeping the compression grid up to date. At the end of the session it may compare the starting version to the one it might use for the compression and, if there isn't a good match, it's easier to just use the second version.

The odds of having to recreate the compression grid over again from scratch are high if the editing included cropping, filtering (like a Gausian blur), and especially resizing. The recreation will indirectly entail what was learned from the Convolutional Neural Networks (CNN).

What's not clear from reading the paper is whether it is describing a proposed enhancement to JPEG with the help of information learned from neural networks or whether neural networks were used in the development of JPEG compression techniques.

I worked with neural networks in the past and I can attest that their creation is not something you can do on the fly. They involve a looping process that repeatedly adjusts the networks assumptions until the output matches or comes close to a desired result. When I used them they could run for hours. With today's computers they will run much faster but they are not running in your camera or on your computer during the creation of a JPEG.