Is it possible....? - DPChallenge Forums

Canon EF-S 10-22mm f/3.5-4.5 USM

05/23/2005 04:19:36 PM · #1

Is it possible to have compression artifacts in a 100% crop of a pic? I'm not trying to complain about one of the comments I got, just trying to understand...
Here's the pic in question.

The little sparkly looking bits are actually in the petal texture. Could this be what was considered compression artifacts? If so, how could you distinguish between the two?
Any help would be appreciated.
Sara

saintaugust

Canon EOS-550D Rebel T2i

05/23/2005 04:21:44 PM · #2

Originally posted by saracat:

Is it possible to have compression artifacts in a 100% crop of a pic? I'm not trying to complain about one of the comments I got, just trying to understand...
Here's the pic in question.

The little sparkly looking bits are actually in the petal texture. Could this be what was considered compression artifacts? If so, how could you distinguish between the two?
Any help would be appreciated.
Sara

i think the artifacts are from sharpening, not compressing

Jeff

srdanz

Canon EOS-5D Mark IV

05/23/2005 04:22:40 PM · #3

Yes, it is. The compression does not refer to the shrinking or expanding the photo, it refers to the jpeg or other algorithm to reduce the size of the image on disk.

I can't tell the difference in your case, though.

Canon EF-S 10-22mm f/3.5-4.5 USM

05/23/2005 06:02:21 PM · #4

Originally posted by srdanz:

The compression does not refer to the shrinking or expanding the photo, it refers to the jpeg or other algorithm to reduce the size of the image on disk.

I'm not sure I understand how this applies ... I'm not well-versed in computerspeak and I may not understand what you mean.
When I referred to this pic being a 100% crop, what I meant was that I took a 640x480 chunk out of the center of the photo (or thereabouts). I didn't resize or scale the image. Was my terminology correct or should I have phrased it differently in my original post?
Also, if I didn't resize or scale the image, would the compression artifacts still be there, or would they only exist as a result of the sharpening? Or should I take your quote to mean that the artifacts result from any alteration to the picture, be it cropping, adjusting contrast, or playing with levels?

Arrgh! Maybe I should just sign up for a class at the local tech college!

Anyway, thanks for the info.
Sara

srdanz

Canon EOS-5D Mark IV

05/23/2005 06:24:46 PM · #5

Compression is orthogonal to resizing. When you resize, you either lose pixels (going from e.g. 1280x960 to 640x480 without cropping) or you gain some pixels (e.g. from 640x480 to 1280x960). Resize algorithms are doing their job and you are looking at the effect on the screen immediately. Same goes for sharpening.

However, when you decide to save your work, and you choose e.g. jpeg, then you are choosing a 'quality' or the compression rate. Your image may be "worth" some 200KB in its 640x480 format, but you may want to reduce it to 150KB for instance to fit the challenge format. The compression will then be seen in the saved image and it will look different from the last image you saw before saving.

ps. if you use 'save for web' you may see what it looks like before actually performing the compression.

Does this help any?

Bear_Music

Samsung Galaxy S24 Ultra

Sony 24-70mm F4 Vario-Tessar T* FE OSS

05/23/2005 06:59:20 PM · #6

Sara,

It's very simple to find out; open you original, open your modified file, zoom in on the same area of each, and compare.

In theory, EVERY time you save a jpg file you are re-compressing, and in theory if you save them often enough the image becoems hopelessly deteriorated. This is why we do all our work in TIFF or PSD or PSP formats, as these do not compress during saves. At full size, one of my PSD images can easily run 60-90 Mb. EASILY. Each layer contains all the information of each other layer. When I am through working, I may have sevral hundreds of megabytes of files on the one image. I take the most highly-evolved one and save it, witha ll its layers and masks and selections, toss the rest, and flatten and save-as the good one, but still as "filename-flat.psd". So I have TWO full-size PSD versions, the one with all the info and the flattened one.

The flattened one I resize, add USM and any other adjustments needed at the smaller size, and save-as "filename-flat-small.psd"; so now I have THREE PSD files, each progressively smaller in size.

Finally, I add my borders and save for web; ending up with 3 PSDs and two JPGs (one unaltered original, one to post at 640 pixels) for every DPC challenge entry.

If you follow a workflow similar to this, and if you do your sharpening at magnification so you can SEE the artifacts forming and fade the sharpenign to avoid them, you should never have a problem.

Sorry for long-windedness, all this may be off-topic...

Robt.

kirbic

Canon EOS R5

Canon RF 24-70mm f/2.8 L IS

05/23/2005 07:07:28 PM · #7

In this case, your file size is only 40,753 bytes (39.8k). When blown up, there are mild compression artifacts visible. They are most apparent in the blocky appearance of smoother areas.
You should shoot for >100k for your submissions, that will greatly reduce the appearance of artifacts. For images with a lot of detail, keep the size as close to the 150k limit as you can (>145k).

05/23/2005 08:10:33 PM · #8

Thanks for the info, ya'll, but I believe it has blown my widdle mind. :)
I'm gonna have to re-read a lot of this stuff over and over before it starts to sink in. :( And here I thought I was smart enough to soak this stuff up pretty quickly! I see it's not quite as intuitive as I'd thought...

More questions:
srdanz, I think I understand what you're saying about the compression due to resizing and saving, but what does 'orthogonal' mean?

bear_music, do you shoot in TIFF or JPEG (or RAW)? I can shoot in TIFF or JPEG; should I start with the pics in TIFF and leave them that way until I'm ready to save for submissions? If so, which file would you use in a validation request: the original TIFF file as it came out of your camera, or the TIFF file saved as jpeg with no alterations? Does saving a TIFF file into a jpeg format cause any changes to the photo, and if so, would it make a difference in a validation request? Or should I shoot jpeg, save a copy, save a copy as a TIFF file, do my adjustments, re-save (or save a copy) as jpeg, then submit? (Sorry that kinda all ran together.)

kirbic, you weren't confusing enough. No questions for you!! :)

General question: Does GIMP have a 'save for web' option that I just can't find?

Thanks in advance,
Sara

Canon EF-S 10-22mm f/3.5-4.5 USM

05/23/2005 11:25:36 PM · #9

bump for the night shift

srdanz

Canon EOS-5D Mark IV

05/24/2005 04:02:37 AM · #10

Night shift reports for duty (I'll sleep in June, I think)

By "orthogonal" I meant unrelated.

As far as shooting mode, I'd recommend raw (or tiff) as it is lossles, i.e. no compression has been made to the image. (And more than that, no processing has been applied to it either.)

This leaves you with larger spectrum of available adjustments you can make to the image. JPEG is smaller in size than raw or tiff, which by common sense means that it is of lesser quality, with less color information per pixel for you to work with.

As far as valid files for validation, it does not matter what's your format, as long as you have saved a copy of your image before you did anything to it, including even opening it up in a photo editing software to view it only. Some SW packages would modify the file without you wanting it/knowing it and you'd be DQed.

Do not convert jpeg to tiff, I can't see the benefit. It is going from less quality to an image with more space to store quality, but it has been already lost - irreversibly.

And yes, any time you convert to jpeg or save a jpeg to jpeg, the compression happens again. The image becomes smaller = less space for the useful information about the colors etc.

Does this leave something unanswered?

To summarize:
take your images in the highest possible mode your camera would allow. (usually raw)
post-process it without resizing until you are happy with the composition, colors, etc.
Save it that way (if you want to make large prints out of it later)
make a copy for DPC portfolio or a challenge. For the challenge, you should aim for larger dimension to be 640, and the size to be highest available yet <150K. Use the save for web to achieve this and adjust the quality. Fine-tune it with the number, don't settle for the discrete levels such as good, better, best.
For the portfolio images, you should still reduce the size to 640, but you do not have to aim for the largest file under 150K. If you want to fit more images, it is usually OK to do the 50-80K photos...

Wow, long post, I hope 25% is useful.

Bear_Music

Samsung Galaxy S24 Ultra

Sony 24-70mm F4 Vario-Tessar T* FE OSS

05/24/2005 04:42:22 AM · #11

I have to disagree with srdanz as to the utility of saving a jpg as a tiff to work on. Regardless of what the original file format was, you want it in a lossless format when you start working on it. Ideally, the native format of the program in which you are working, but tiff will do as well. There's a simple reason for this, beyond the lossless aspect: jpg images cannot retain layers and selections. Tiffs and native formats can. You want your working copy to vontain all the adjustment layers and all the selections you made and saved within itself, so you can go back and tinker with the image as the need arises.

When you're ready to put it on the web, you flatten that image and save-as (again lossless) so you have a full-size version in a single layer. Then you resize that, and save-as again in the lossless format, so you have a 640-pixel, flat, lossless image; that's your base image to go to the web from. If you intend to use borders, you should calculate the border size before resizing, and subtract twice that from 640 to get your final size. You sharpen this a little, if it needs it, then add border if you're using one. Then you save-for-web or, if you lack that feature, save-as a jpg and manually set the slider to a quality level that brings the filesize to around 140k. The advantage of save for web is that it strips the "tags" off the image, as you don't need 'em for the wb. If the tags are appended, your nominal 140k image will be closer to 150k.

Sara, to answer your question, when I'm shooting for DPC challenges I usually shoot jpg at the fine level; it's more than good enough for 640 pixel shots. If what I'm shooting looks like stuff I'd want to print and hang eventually, I go to TIFF. I don't use RAW on the 5700, though I would if I had a dSLR.

Robt.

edited to clarify resizing for border

Message edited by author 2005-05-24 08:54:17.

buzzmom

Panasonic DMC-FZ20

05/24/2005 07:59:22 AM · #12

wow you can teach an old dog new tricks...i now understand what happened to my apple submission....lesson learned.....

thanks guys... (maybe when i smarten up bear will add me to his dpc dinner) list....

glad2badad

Apple iPhone 16 Pro

Lensbaby Sweet 35 Optic

05/24/2005 08:10:13 AM · #13

Robert - Great workflow advice. I'm re-reading to check against what I use and I've found some useful tips as always from your posts. One question however, wouldn't you add the border BEFORE resizing to 640? If you add it after the image is too large and needs additional resizing for the challenge.

Take care! ;^)

Bear_Music

Samsung Galaxy S24 Ultra

Sony 24-70mm F4 Vario-Tessar T* FE OSS

05/24/2005 08:50:54 AM · #14

Originally posted by glad2badad:

Nope, and here's why: the monitors can't resolve less than a pixel, obviously. So if I make, say, an 8-pixel border at full size (giving me 16 pixels of border right and left, 8 on each side) and then downsize to 640 pixels, the downsized border might be, say, 7 pixels total, or nominally 3.5 pixels per side, and it can't show that, so it gets lopsided. What I do is calculate in advance how much to allow for the border before resizing; so if I want 5 pixels of border all around (2 white and 3 black, say) I resize to 630 then add the white and black broders.

Also, and this can be important, typically I need to do an extra pass of USM after I resize my image. If I have a contrasting border already on it, the USM will sometimes create the ghost of an artifact wherever the contrast is sufficient. So I downsize, sharpen, then add border as last step before saving for web.

Robt.

Ah, I see the confusion; I did specify "640" in the earlier post. I'll edit that.

Message edited by author 2005-05-24 08:52:17.