Image degradation with CS2

[D. King]

I have been reading that it is preferable to make various adjustments to raw images using the Adobe Camera Raw Plug-In because making some of these same adjustments in PS causes the image to degrade.  

I have not been able to find any info on the kinds of adjustments and the degree of degradations, if any, they cause in PS.

As a possibly extreme example, using the various "transform" commands -- stretching, compressing the image to fix perspective etc -- seem like a likely candidate.  

Is there a source of this kind of information?

Thanks.

[BernardLanguillier]

I have been reading that it is preferable to make various adjustments to raw images using the Adobe Camera Raw Plug-In because making some of these same adjustments in PS causes the image to degrade. 

I have not been able to find any info on the kinds of adjustments and the degree of degradations, if any, they cause in PS.

As a possibly extreme example, using the various "transform" commands -- stretching, compressing the image to fix perspective etc -- seem like a likely candidate. 

Is there a source of this kind of information?

Thanks.
[a href=\"index.php?act=findpost&pid=87775\"][{POST_SNAPBACK}][/a]

Several books on PS have sections on this. I'd have to check which ones.

General considerations

Simply put, there are at least 3 types of degradation:

1. Generally speaking, increase of what I call color entropy in the image. Pixels that initially were carrying different color information end up having the same. The macro effect of this is typically banding and a softer look of the image. I call this entropic because it is a non reversible process that results in the image becoming more homogoneous. You lose some informational content. Curves and such modifications typically have this effect.

You can usually check the occurance of this by looking at the histogram of an image. It gets to have a comb like look that means that some RGB values are not used anymore although they originally were.

2. An important sub-set of the above is clipping of the colors. When applying a modification to an image that causes one or several RGB channels to hit the dark or bright ends of the spectrum, pixels that initially had different color information all crash into 0,0,0 or 255,255,255. This is especially ugly on the bright side since you end up with whole areas of the image that are uniformly white. Ugly.

3. Decrease of sharpness. This is indeed the result of all "geometrical" modifications of the image including modification of pixel count.

How does working on RAW impact?

It seems that at least type one degradations are a lot less severe when done before RAW conversion. This would have to do with:

1. The fact that the modifications are then done on the linear RAW data before application of a curve.

2. The fact that cameras using a Bayer pattern sensor need demoisaicing to produce a viewable RGB image. Compared to a Foveon like device, Bayer pattern have less color information. The demoisaicing algo is therefore related to this color information or loss thereof also.

I have never really looked into the math of this and have decided to believe those who have and reached these conclusions. Empirically, you can do some tests that do back up these claims.

Hope that it helps.

Regards,
Bernard

[Dale_Cotton]

Bernard-san wrote:

increase of what I call color entropy in the image. Pixels that initially were carrying different color information end up having the same.

Whoa! Beautiful analogy!

Add to your list of (potential?) benefits of raw: ALL changes - Bayer, contrast curve, WB, etc. - can be batched together into one arithmetic transform before being applied to the RGB data - which should save on rounding errors. In Photoshop or other editor, one would use Curves to change contrast, then Hue/Sat to change colours, etc. each as a separate assault on the colour numbers of each pixel. (Does 16-bit-ness absolve us of this crime? I'm not number-savvy enough to comment: but surely the effect increases the further one descends into the lower zones of shadow-hood?)

[BernardLanguillier]

In Photoshop or other editor, one would use Curves to change contrast, then Hue/Sat to change colours, etc. each as a separate assault on the colour numbers of each pixel.
[a href=\"index.php?act=findpost&pid=87781\"][{POST_SNAPBACK}][/a]

Dale san,

That is true except if you use Adjustement layers in PS. When using adjustement layers, I believe that the total transformation is computed before being applied to the underlying pixels.

Regards,
Bernard

[Jonathan Wienke]

Color entropy can be reduced to negligible levels by doing all edits in 16-bit mode. That is why doing all major edits in the RAW converter is best if you use 8-bit mode; the transforms introducing the most color entropy are done inside the RAW converter in 16-bit mode before the result is rounded to the nearest 8 bits. But overall, it is best to do everything in 16-bit mode.