Gimp vs Photoshop

External Since: Oct 04, 2005 Posts: 833

Bill Tuthill wrote:
> Floyd L. Davidson <floyd.DeleteThis@apaflo.com> wrote:
>>>>> Moreover, even when closely reproducing JPEG quality settings,
>>>>> Photoshop JPEG is bloated, almost 2x the size of GIMP saves.
>> Lets be clear that attributions have been trimmed and
>> the above is *not* anything that I said.

Sorry Floyd, I did mean to confuse attributions.
>>
>>>> Yikes.
>> That was my comment!
>
> I was the "2x bloated" poster, and I stand by my statement.
> Here is some proof; the PS is 57% larger without EXIF or preview
> (and preview must be manually disabled):
> http://photo.net/bboard/q-and-a-fetch-msg?msg_id=00MJfw

Perhaps the following is the reason I haven't noticed the effect
you cite: I have rarely used jpeg quality settings below
about 50% because I find the degradation too great, regardless
of the program used. The above web page is an example
of poor images. It also seems you have not done a straight
comparison, as the photoshop ones have had a lot of
other processing done, and that changes the detail and
compression results.
>
>> You are essentially confirming that by saying "save for
>> web", which indicates to me that arbitrary choice of
>> quality is not possible, and instead quality must be
>> selected from a set of specific uses. Not to mention
>> the comment about chroma sampling...
>
> I just tried PS CS2's Save For Web. I could not make it write
> 2x1 chroma subsampling. It did write 1x1 at quality levels
> down to somewhere between 50 and 60, where it switched to 2x2.

And where image quality drops to really bad.
>
> CS2's "Powered by ImageReady" dialog allows a wide choice
> of quality levels from 0 to 100, but they have no relationship
> that I can discern to IJG quality levels. At ImageReady Q 50,
> I was able to duplicate IJG 80 2x2 at about the same file size.
>
> So the interface is basically broken for editing JPEG images
> from DSLR and modern digicams, which use 2x1 chroma subsampling.

I disagree that its broken, and I wouldn't use chroma subsampling
if it were available, but that's just me. When I do choose to
save jpegs, I generally use high settings, like 90% and greater.
Disk space is cheap and getting cheaper by the day.
When using reasonably high quality settings, the differences
between converters is not much different in my experience.

Roger

External Since: Oct 04, 2005 Posts: 833

Dave Banks wrote:

> When a program, such as PhotoShop, uses bicubic algorithms on all their image
> manipulation tools like resizing, rotations, leveling, skewing, perspective
> corrections, lens-distortion tools, layer manipulations, and any other tool that
> requires that the pixel data be resampled and shifted elsewhere (which would
> also apply to their cloning and healing brush-tools because they don't usually
> clone on pixel boundaries), then there will be a softening and blurring of that
> image data. This is why using the old method of bicubic, or even poorer
> algorithms like bilinear interpolation, as the core process to all features in
> any editor is such a phenomenal setback.

I have experience with a lot of interpolation algorithms and NONE
are perfect. ALL have artifacts in my experience (scientific
data as well as consumer imaging). Bicubic interpolation
is relatively robust for a wide variety of situations, of which
digital camera image processing is quite diverse, so it's choice
is reasonable. It is unfortunate that Adobe hasn't given other
choices, however.

> It's been that way since day-one with PhotoShop. You would think that the vast
> numbers of these supposedly highly intelligent, keen-eyed, and discerning
> photo-editing "Pros" who rely on PhotoShop all their lives and constantly advise
> everyone else follow in their ways would have noticed all this by now. It just
> goes to prove that anyone can claim to be anything but they still don't know
> jack. I guess this also proves that 20 million people, even (cough) "Pros",
> _can_ all be wrong. How about that. Smile

This rant against bicubic interpolation is minor considering
many of the other larger problems with photoshop, of which
many people seem unaware.

1) 15-bit processing. It was fine in film days, but now with
high S/N digital camera data, it can show problems.
Photoshop needs to fix this, including doing 32-bit internal
processing.

2) Photoshop was born so long ago when computer were very
slow, many algorithms are additive approximations, but should
be multiplicative (e.g. contrast, saturation and hue adjustments,
for example). I don't use these tools in photoshop or use them
in other programs that do the correct math. If the corrections
are small, these aren't big problems, but as you need to do
greater corrections, it can become horrible (try sliding the
saturation slider all the way for example).

3) Real sharpening filters are needed, like Richardson-Lucy
deconvolution.

> Try to never depend on the advice of others, no matter who they claim to be, or
> who others claim they are. Test things for yourself and you too will uncover the
> outlandish deceptions being promoted by all self-appointed authority figures
> with a personal agenda that surround you daily. This news-group is filled to the
> brim with a vast surplus of those fools.

It isn't hard to figure out who has credibility in the
newsgroup by watching posts for a while. Unfortunately,
many very helpful people have left (the regulars can
probably name a few).

Roger