JPEG Degradation

[arnold_theisen1]

It is often stated here and on other forums that each time you save

a JPEG image it becomes slightly degraded. Yet, I read Ken

Rockwell's eloquent defense of JPEG

(http://www.kenrockwell.com/tech/raw.htm) and I wonder just how much

do we have to worry about this phenomenon. Rockwell doesn't address

this issue in his essay. I would like to ask him, but he requires a

tribute to answer emails. Right now his going rate is $2.00 a

minute, which is more than I can afford.

 

I tried a little test of my own. I scanned a good slide at 4,000 dpi

to both JPEG and TIFF formats. This gave me an image of 3,470 X 5263

pixels. The JPEG file size was 11.6 mb, and the TIFF file size was

27.2 mb. I opened the JPEG file in Photoshop Elements and saved it

20 times, each time using the save-as feature and renaming the file

each time. I then compared the 20th save with the original and can

see no difference on my monitor even at 200% magnification. It also

looks as good as the TIFF file made at the same time. I noticed that

the JPEG file increased from 11.6 mb to 13.1 mb after the first

save, then to 13.2 mb after the 10th save and then to 13.3 mb on the

20th save. Don't know what accounts for that.

 

I'm not a techie, so I don't know if I've proven anything here or

not, but I really would like to know if the JPEG degradation issue

is anything to worry about. Does it require hundreds or thousands of

saves to make a difference? Any insight will be appreciated. I'm

looking at the Nikon D70 and I would rather not mess with raw files,

yet the only choices are raw or JPEG. Evidently the D70 won't

capture TIFF images.

 

How about if I take an original JPEG and immediately convert it to

TIFF? Does that then protect against any further degradation?

[ellis_vener_photography]

Jpegs only degrade if you do something to the file itself and then resave as a jpeg.

 

jpegs are also only allowed by the JPEG (but not the JPEG 2000) standard to only be 8 bit

color depth (256 shades or steps) per Red, Green and Blue channel. NEF files are 12 bit

per channel files (4096 steps per R,G &B channel.) the extra shades are very useful when

editing an amge file because virtually every edit you make on the file throws away some

set amount of data. S oeven if you can only print at 8 bit color depth, you preserve enough

data to avoid banding and other problems that show uo when you have thrown away

enough data out of an 8 bit original file.

 

Converting to a TIFF immediately is a good stategy if you are planning on doing some

digital darkroom (AKA "editing") work on the image.

 

A NEF file from a D70 will have all of image capture that a TIFF file would contain but will

be much smaller than a TIFF. The reason it is smaller is that won't have a lot of the

processed image data that is included in the TIFF format. For one thng there is no color

interpolation.

 

Color interpolatiion? what the hell is that?

 

Glad you asked.

 

(Now for the big digression that roughly explains color interpolation and whay almost any

camera you buy interpolates color. Cue the heraldic trumpets)

 

All digital capture systems you see in any 1 shot digital camera interpolate color . Every

pixel in a sensor has either a green, red or blue filter over the individual pixel in the

sensor. What thappens when the iamge is proceseed from the raw capture is that the

processing software looks at the surrounding pixels and interpolates that data with the

data for that pixel to get a full set of R,G, & B values for that pixel. The one exceptio nto

this seems to be the Foveon chip. The only way around this color interpolation is a multi-

shot method which is used in the higher end digital backs made for medium and large

format cameras. Multi-shot works by phsically moving the entire sensor in either 1/2 or

full pixel increments s ot hat each point in the photograph is photographed four or 16

times. The obvious drawback of this is that the subject can't move, the camera can't

move and the light can't change between each exposure. the benefit is is much higher

color quality and in the case of backs used in 16 shot mode, much higher resolution.

[alexdi]

Ken Rockwell's opinion is a perfect solution for Ken Rockwell. Anyone who isn't Ken Rockwell would benefit from forming his own.

 

There are three advantages to RAW over JPEG: A) more adjustment latitude, B) the ability to compensate after the fact for horrific lighting, and C) the chance that someone will release a significantly better RAW converter. If these outweigh the increased hardware demands and postprocessing time, RAW's your ticket.

 

DI

[beno_t_marchal]

<p>Ken's point, if I understand it correctly, is that JPEG is appropriate for prolific shooters

who do not have the time (or the willingness) to edit their shots. He makes it clear that he

fells into this category. He suggests RAW may be appropriate when you have more free

time and you intend to edit the image.

<p>I think that's a sound opinion: essentially Ken is telling you to study your shooting

style and adopt the tools that work best for you.

<p>You will hear many different views on photo.net (and elsewhere on the Internet).

People may be passionate about a workflow but remember that a good

workflow reflects on somebody's shooting style and personality.

<p>It is likely my clothes would not fit you... does that make them bad clothes? No but

they reflect my style and personality.

<p>Likewise for photography. One may prefer studio lights, one may prefer natural light.

Does that make one better or worse? It's the image that matters.

<p>Personally I am interested in still-life, landscape and I recently got interested in

portraiture. I like to take my time, I like to control my lighting and I don't shy at post-

editing. I have found RAW gives me that tiny bit of additional control that I crave for.

<p>I don't limit myself to RAW though. First I shoot my test images (digital Polaroid) in

JPEG: it's faster to download and the card does not fill up so quickly.

<p>Second being known as the photo guy by my friends and familly, I am often asked to

shoot action. For action photography under stable lights, I have found JPEG works best. It's

faster, takes less space and I don't need to bring a laptop with me. If the light is not stable

then I might use RAW because I may have to post-process the photos.

<p>You don't have to adopt mine or Ken's workflow. But you may want to find a workflow

that works for you.

<p><i>I opened the JPEG file in Photoshop Elements and saved it 20 times, each time

using the save-as feature and renaming the file each time. I then compared the 20th save

with the original and can see no difference on my monitor even at 200%

magnification.</i>

<p>Unfortunately the test is biased and not representative of the degradations you would

see in real use. I try not to be technical but if you recompress the exact same image at the

same compression level, using the same JPEG implementation (what you did) then the

algorithm will introduce the same artefacts at the same position... no visible

degradation beyond the first save.

<p><i>How about if I take an original JPEG and immediately convert it to TIFF? Does that

then protect against any further degradation?</i>

<p>If you don't plan to edit the image, don't bother. You don't loose anything by

opening a JPEG a thousand times. It's only when you edit and save. So don't bother saving

it in TIFF until you edit.

<p>--ben<br><a href="http://www.marchal.com">marchal.com</a>

[johnb]

Only compression degrades jpeg files, just tell software to not compress the jpeg and there is no degration. Rockwell is right but all the Raw people love to argue there point. Thats what I noticed!

[steve_dunn2]

<cite>Only compression degrades jpeg files, just tell software to not compress the jpeg and there is no degration.</cite>

 

<p>Assuming that we're talking about traditional JPEG here, not JPEG2000, this is highly unlikely. <a href="http://www.jpeg.org/faq.phtml?action=show_answer&question_id=q404fa5e29aeb6" target="_blank">The original JPEG specification does allow for lossless compression, but very few programs actually support this.</a> If you want to test your favourite program, save the same file as a compressed TIFF (TIFF uses lossless compression) at 8-bit colour depth, and as a JPEG at whatever setting you think would not be compressed. If the JPEG is significantly smaller than the TIFF, your software is using lossy compression (the usual type for JPEGs).</p>

 

<p>Back to the original topic - <a href="http://www.jpeg.org/faq.phtml?action=show_answer&question_id=q402c4ff3598f4" target="_blank">this question is answered in the JPEG FAQ</a>.</p>

[bill_tuthill]

It's just dumb to have a RAW-capable camera and shoot in JPEG mode.

Someday you might want to edit or enlarge one of those images.

If your digicam shoots only JPEG, or somebody gives you JPEG, I used

to think it useful to convert to TIFF. But lately I've learned

less-destructive editing methods, including how to deal with 8x8 block

boundaries, and preserve quality level and chroma subsampling,

using GIMP and jpegdump. P.S. Nice to see so much sound JPEG

knowledge here.

[ellis_vener_photography]

<I>It's just dumb to have a RAW-capable camera and shoot in JPEG mode.

</I><P> That is a little harsh and it is not exactly true either. It really depends on what

you are shooting, and what you think you

are going to use the photos for, as well as the circumstances you are working in. As an

example of the latter: you have been shooting all day in NEF mode and are running out of

media. Would it then be better not to shoot at all or to shoot jPEGs?<P>

 

But I do think it is dumb to shoot anything but high (AKA "Fine") quality / large resolution

JPEGs if you can't or choose not to shot RAW/NEF files.

[arnold_theisen1]

Thanks for all the good advice and suggestions. It looks like I'll need to shoot in raw mode at least part of the time. But now I'm better equipped to determine when to use raw and when to use JPEG.

[jespdj]

Bill Tuthill Photo.net Hero, aug 19, 2004; 01:34 p.m. wrote:

<p><i>It's just dumb to have a RAW-capable camera and shoot in JPEG mode.</i>

<p>Yay! Let's have another JPEG-vs-RAW war!

[jon_austin]

I just scanned the replies; didn't read them all carefully, but wanted to toss in that if you have a JPEG image open in your editor, and save it over and over countless times, you won't observe any degradation. However, the cycle of opening, saving and closing, reopening, saving and closing, etc., will indeed induce generational degradation when compared to earlier generations of the image.

[bill_tuthill]

Ellis Vener is right of course, but I still believe JPEG is overhyped

(have you ever tried to color-edit even the highest quality JPEGs?)

because the camera companies don't want to pay higher royalties for

something better.

<P>

Jon, I just tested ImageMagick to see JPEG generational differences.

After the first TIFF to JPEG save, the second resave was slightly

different near the beginning and at the end. The third and successive

resaves were slightly different only at the end, repeatedly switching

back and forth. Perhaps a rounding error in DCT calculations?

<P>

When people say "I don't notice degradation in high quality JPEG" what

they really mean is that "JPEG was so well designed that it fools me."

Try HSV mods <A HREF="http://www.photo.net/bboard/q-and-a-fetch-msg?msg_id=009Heq">as

I did in this thread</A> and you will see how easily fooled you were.

Or try an image "diff": more than half the pixels change between

TIFF and JPEG even at quality 99 and 1x1 chroma subsampling.

Indeed, I believe JPEG is one of the the 7 wonders of modern world,

but primarily as a web presentation technology.

[beno_t_marchal]

<p>That depends on the software. For example, Photoshop compress again (degrades), PhotoMechanic does not.</p>

[bill_tuthill]

Benoit, it is "interesting" that in the 6 years since this thread was last updated, I have switched entirely to JPEG editing with GIMP, which has a "save at original settings" option. As a JPEG editor it works a lot better than Photoshop. I am not wealthy enough for RAW, Lightroom, etc. PhotoMechanic seems interesting however it is not as full-featured as Photoshop.

[zedex]

<p>Does renaming or adding IPTC info constitute "editing"? It does in my mind, therefore a JPEG would suffer degradation as the file would be compressed after opening to add the IPTC info. I can't find a 100% clear cut answer with regard to IMatch.</p>

<p>With XnView adding IPTC does not degrade the JPEG as long as the info is saved with the "Write" or "write all selected." It does with the "apply."</p>

<p>I add all IPTC comments with XnView because I know what I'm doing with the software. I can keep doing that, but it is a pain to have to update IMatch every time I open it.</p>

[bill_tuthill]

Bruce, if XnView is able to insert IPTC tags without changing the JPEG encoding tables, this should be a lossless operation. I believe <b>exiftool</b> changes EXIF tags without recoding JPEG. However Photoshop or GIMP will recode the JPEG. I did a recent experiment with Photoshop, which does not have an option for 2x1 chroma subsampling, although this is what almost all digital cameras produce. Fortunately the extra lossiness of going from 2x1 to 1x1, which Photoshop does at higher Q values, is not highly visible. However changing the Q values is very destructive if you look closely.

[zedex]

<p>Hello Bill</p>

<p>Thank you and the others too. I have it all sorted and in case anyone else is interested, Irfan, XnView and iMatch do not recode when IPTC is entered and the file(s) closed. Therefore it is safe to add the IPTC info in dribs and drabs. This is of importance to me as I am more of an archivist than a photographer and started with the family photos, thousands of them. I believe that in order for them to be relevant to the younger and future generations some explanations and identifications where necessary. IPTC seemed the best way to do it. Thanks again.</p>