Post Reply 

Two questions...

Nov 3, 2013, 09:30
Post: #1
Two questions...
1. After I use Gimp for awhile, it seems to get "all clotty and rancid", plugged up, slow, especially if I do something that takes a long time like Foreground Selection for example. Sometimes I can't tell if it's doing anything at all or just frozen. If I close it and start again, this seems to clear up the problem. When I'm working on photos, I leave it open after the first one so the start-up for the next one is quicker. I usually close whatever image I've worked on, but just leave the program open. Does the program have a "cache" or something I can clear and still leave it open or should I just suck it up and close it every time?

2. This is a general question about image degradation. I read that .jpgs degrade some every time you open and close them. Since I do mostly photo stuff, this is the format I use the most just for the sake of convenience. How much of a problem is this? Do some tools degrade the image more than others? Does the order in which I use tools or the number of times I use them have any effect? Sometimes I notice posterization, is this because I'm just working the image too much? When I open an image a second time, should I open the .exf file instead of the .jpg? What about 8bit vs. 16bit? Does this really make that much difference? Is there a workflow that's less degrading... besides working in my Little Bo Peep outfit, I mean.

Don't forget to set your clocks back. Or forward. Or whatever it is.
Find all posts by this user
Quote this message in a reply
Nov 4, 2013, 01:23
Post: #2
RE: Two questions...
Btavern to answer your #1 question (Of answered quite nicely #2 & #1)

#1
You can reduce the number of undo levels(can waste memory) to 2, 1, or maybe 0:
http://docs.gimp.org/en/gimp-pimping.htm...nvironment

Undo is like a cache but you can save your original project as a back up should you need to revert to it / .XCF (save as...) file then just have another/new file to work on & turn down/off the undo states.

This to me, has been known to clog down my gimp, but I just clear the undo states manually in the undo dialog window every 5, 10, 15 minutes if I'm doing a lot of editing to the photo/project.

#2 Just opening and closing a jpeg won't degrade it, its when you resave it with high or lower quality than the original. There are some edits you can get away with that won't degrade the image on save much if any. You can always undo before you save.
*If you're really worried about losing quality, be sure to work on a copy image(best) or layer underneath (save your original raw photos in a folder somewhere)

If you know you won't accidentally edit the bottom layer(and the image file size is around 1 to 2 mbs or less), I'll duplicate the image as a new layer and work on that one, then save an incomplete project as a .xcf, then export as jpeg when done/final.

"In order to succeed, your desire for success should be greater than fear of failure." BC
[Image: 29uocaf.png]
iFS007 tut vids on YT ~ auto color B&W landscape tut
Visit this user's website Find all posts by this user
Quote this message in a reply
Nov 4, 2013, 05:46
Post: #3
RE: Two questions...
Thanks you guys. I have just a couple more questions... naturally.

1. How far back does the history go? It seems to go back at least a month and I use Gimp almost every day. I just deleted all of it and we'll see if that makes a difference. I hope that wasn't a tragic error.

2. Can either of you--or anyone else--recommend some settings for:
Maximum Undo Memory (now at 10mb)
Tile Cache Size (now at 644mb)
New Image Size (now at 128mb)
Would lowering those sizes reduce the impact Gimp has on my system memory? I pretty much only use it for photos and I don't think any of them get above, say, 20mb.

2. Ofnuts: The digiKam suggestion is much appreciated. I've been using Faststone Image Viewer for some of those functions, but just looking at the digiKam site, it seems a lot more useful. Being able to do the global color stuff in another program would be a big help. The main thing I use Gimp for is color stuff and a pseudo-high pass sharpening thing I use that involves Grain Extract, Grain Merge. I don't like USM and am not skilled enough at G'MIC to really use it effectively yet. Oh yeah, I use Threshold for some selections now, thanks to iFor.

Also knowing about maybe using scaling to mitigate the effects of global adjustments is interesting, thanks. And NOTthanks for the monotonic link. Before I was done, I was six pages away, looking at a wiki on Riemann-Stieltjes integrals and Darboux sums with Greek symbols and blood leaking out my eyeballs. I'll get you for that, it made me miss the Girly-Man Ice Dancing Finals on NBC. Seriously, the digiKam thing was most excellent, thanks.

iForStyle007: I'd forgotten entirely about the whole "pimping your gimp" stuff in docs. That was a big help. And your workflow suggestions are always the best. If you hadn't shown your cruel side in that Moltes thing, I'd wanna be your bitch. What do you set those environment variable I mentioned above at? Thanks for the help.

I got an email today the indicated that Gimp 2.8.8 is out. Hmmm... And has anybody looked at LightZone? It's free, but it's vector-based. How much difference does that make for photos? It's got some cool zone adjustments and uses a sort of layers thing like Lightroom or Photoshop, I think.
Find all posts by this user
Quote this message in a reply
Nov 4, 2013, 07:21
Post: #4
RE: Two questions...
Wow, thats a mouthful of text, lol I'll try to reply to all your questions but I only like to type alot when making tutorials/explanations so this might be short.

Quote:1. How far back does the history go? It seems to go back at least a month and I use Gimp almost every day. I just deleted all of it and we'll see if that makes a difference. I hope that wasn't a tragic error.
Woa, a month! Gimp doesn't have a global undo state, unless you work in the same file when you reopen. The undo history in gimp can go all the way back to the start of your project if you haven't cleared it. To my knowing there isn't a program history, just the undo history states.
Quote:2. Can either of you--or anyone else--recommend some settings for:
Maximum Undo Memory (now at 10mb)
Tile Cache Size (now at 644mb)
New Image Size (now at 128mb)
Would lowering those sizes reduce the impact Gimp has on my system memory? I pretty much only use it for photos and I don't think any of them get above, say, 20mb.
First of all, what are your computer specs? Amount of RAM, free hard drive space? Amount of free space on your drive?
- open your system task manager, click performance: how much ram is used.
- then open just gimp and check the performance tab again. (this tells you how much gimp is taking / is left)
-- Last, open your average project starting point before you start editing(imported photos, layers, etc) then check the performance tab again to see how much ram you have left.
Then we can determine appropriate settings for your computer.
-I would suggest turning down your new image size to 20mb or less to save space.
Quote:iForStyle007: I'd forgotten entirely about the whole "pimping your gimp" stuff in docs. That was a big help. And your workflow suggestions are always the best.
Thanks Btavern I try hahaha ^_^, now to the second half not quoted, you must understand, "m" has the maturity of a 5/10 yearold in a 40-something's body. I try to be tolerant but you have to draw the line somewhere. I'm not a mean person, just review all his post/threads if you don't believe me.
Visit this user's website Find all posts by this user
Quote this message in a reply
Nov 5, 2013, 06:25
Post: #5
RE: Two questions...
Thanks you guys.
Find all posts by this user
Quote this message in a reply
Nov 5, 2013, 13:41 (This post was last modified: Nov 5, 2013 14:02 by foxmuldr.)
Post: #6
RE: Two questions...
(Nov 3, 2013 12:27)ofnuts Wrote:  Jpeg-induced degradation on edit is a recurring topic... It is only partially true. The JPEG algorithm is stable, if you save the image with the very same parameters you used previously, you end up saving the same data and there is no degradation.

This is not true. JPEG is lossy. It always loses color data. Anyone can see the difference through a simple test.

There are two parts, and the test is easy to reproduce:

(1) Grab an original image using Alt+PrtScrn on some window running on your desktop. Trim it down to the desired rectangle. Save as JPEG to test1.jpg with 100% and all of the highest settings.

(2) Load The test1.jpg image in with GIMP. No edits, just saved and re-loaded from the "Open recent" list. Select-all, copy, go back to the original window where you did the original test1.jpg save (the data in GIMP's memory is still 100% color data not affected by the JPEG export), and paste. Before merging down, change the paste layer mode to "difference," then merge down. Initially it will all look black. To see the color loss, go in to brightness / contrast settings and adjust the brightness and contrast values both to their maximum setting. When you do, you'll see the many, many pixels that were changed due to JPEG's lossy algorithm. Everywhere you see something that's not the general dark gray background color, that was color data that was lost by the JPEG algorithm, even on 100 with the best settings.

If you repeat this exact same process using PNG or BMP, you'll see it is all one color, indicating no color loss, as these formats are both lossless.

This always happens with JPEG because it is lossy. To get lossless encoding you need PNG, BMP, or some other format which does not lose color data no matter how many saves. And with PNG's built-in compression, there's really no reason to use any other format for exported files, unless the target system does not support it.

The images below are a capture of an early draft of the above text from my browser. :-)
(1) test1.jpg -- the image saved as JPEG using 100%, 4:4:4, and Floating Point
(2) test1.png -- the image saved as PNG using defaults
(3) test2.png -- the difference between the two with brightness and contrast each set to 127.

Note: Instead of going in to adjust the brightness and contrast, you can also see the areas where the color data was lost by going into paint mode, selecting a white color, setting the threshold to 0, and then using a minesweeper approach, start painting here and there around the image. You'll see where it stops indicating where the color data is different than the pixel you initially painted.


Attached File(s) Thumbnail(s)
           
Find all posts by this user
Quote this message in a reply
Nov 5, 2013, 14:08 (This post was last modified: Nov 5, 2013 14:09 by iForStyle007.)
Post: #7
RE: Two questions...
(Nov 5, 2013 13:41)foxmuldr Wrote:  
(Nov 3, 2013 12:27)ofnuts Wrote:  Jpeg-induced degradation on edit is a recurring topic... It is only partially true. The JPEG algorithm is stable, if you save the image with the very same parameters you used previously, you end up saving the same data and there is no degradation.

This is not true. JPEG is lossy. It always loses color data. Anyone can see the difference through a simple test...
The images below are a capture of the above text from my browser.
(1) test1.jpg -- the image saved as JPEG using 100%, 4:4:4, and Floating Point
(2) test1.png -- the image saved as PNG using defaults
(3) test2.png -- the difference between the two with brightness and contrast each set to 127.
more specifically this vvvvvvvvvvv
Quote:JPEG using 100%, 4:4:4, and Floating Point

That isn't the default of the jpeg photo (original image) by turning the quality up to 100, & saving, even if you don't make any edits (or dramatic edits) you'll introduce new jpeg artifacts because its doing some resampling.

Checking the "[]Use quality settings from original image", checkbox reduces this resampling.
Subsampling at "best" is recommended
The default setting "quality:" for my jpeg is usually 90%, and DCT method default is Integer.

So since jpeg has the potential to degrade, for that very reason, its a good idea to either make a copy in explorer or the folder of that image (right-click copy, paste) or save a .png copy of it as a backup, since png is lossless.

Perhaps if you tested with the original settings of the jpeg you import (screenshot or otherwise) it would further concrete your theory.
Notice also Of said jpeg degradation it was PARTIALLY true, meaning not fully true.

Cheers!!
Visit this user's website Find all posts by this user
Quote this message in a reply
Nov 5, 2013, 15:00 (This post was last modified: Nov 5, 2013 15:05 by foxmuldr.)
Post: #8
RE: Two questions...
(Nov 5, 2013 14:08)iForStyle007 Wrote:  Checking the "[]Use quality settings from original image", checkbox reduces this resampling. Subsampling at "best" is recommended. The default setting "quality:" for my jpeg is usually 90%, and DCT method default is Integer.


Here are the differences when using defaults (90% DCT Integer).

Note: I chose 100% and Floating Point because it does the minimal amount of resampling, producing the image closest to the original. Anything below 100% will always be worse whenever multiple colors are involved. JPEG is a de facto lossy algorithm. No matter what settings you use, it will always lose color data. The only exceptions are when the image has one color (or possibly on properly aligned blocks of single colors).

Update: I looked at the Wikipedia article on JPEG and they say it does support a lossless mode as well, stating, "The compression method is usually lossy, meaning that some original image information is lost and cannot be restored, possibly affecting image quality. There is an optional lossless mode defined in the JPEG standard; however, this mode is not widely supported in products."


Attached File(s) Thumbnail(s)
           
Find all posts by this user
Quote this message in a reply
Nov 5, 2013, 15:37
Post: #9
RE: Two questions...
Ok

I hate to burst your bubble, but its kinda difficult to see much difference between both your #1 & #2(default) pictures as your example is mostly text-based. Text based jpeg compression can be quite high without loss of color data because there is little to no color.

The #3 photo did see an increase in degradation, slightly, still the example is not optimal.

You've already concreted your evidence (just now) Although, I don't know why your so hard-pressed on proving a partial truth completely false when its 50/50

As long as you don't go crazy with editing or compression you will have little to no degradation. You kind of countered your claims when you looked up jpeg on wiki.

Quote: possibly affecting image quality. There is an optional lossless mode defined in the JPEG standard; however, this mode is not widely supported in products."
keyword "possibly" going in circles back to partial, maybe, the list goes on.

I'm done, you can further this debate with Of if he's so inclined. It is an interesting controversy though I'll give you that.
jeers Huh Bycolorselect Huh
Visit this user's website Find all posts by this user
Quote this message in a reply
Nov 5, 2013, 16:09
Post: #10
RE: Two questions...
[quote='iForStyle007' pid='22570' dateline='1383665867']As long as you don't go crazy with editing or compression you will have little to no degradation. You kind of countered your claims when you looked up jpeg on wiki.[quote]

The lossless definitions are relatively new, and not supported in GIMP. And with each successive save and load, it's like making a copy of a copy of a copy, it loses information.

The reason you can' see much difference between the JPG and the PNG is the reason why JPG was invented. The human eye ignores certain artifacts, and therefore it was created to take advantage of that. By removing certain things, our eyes will tell little or no difference between two images. However, when you continually load and re-save you'll see the degradation goes up notably.

Here is a graphical image to demonstrate the differences, which are far greater than text. Here is the original image: http://www.aafida.com/en/wp-content/uplo...-38956.jpg

The attached images are loaded once into GIMP, twice, three times, and so on, with their difference maps shown.

Image descriptions:
(1) Original
(2) Differences between first JPEG 90% Integer save and original
(3) Differences between second JPEG/90/I save and original
(4) Differences between first JPEG/90/I and second JPEG/90/I
(5) Differences between third JPEG/90/I save and original
(6) Differences between second JPEG/90/I and third JPEG/90/I.

Each one shows differences, meaning at each stage color data was lost. If someone wants to write a script to repeat this from the command line several times, you'll begin to see that enough color information is lost through what would be successive edits that it is not a form to use. The only reason it's not a big deal in most cases is because the sampling data (pixels per inch of viewing screen) is usually large enough to overcome the loss. Still, the loss is there, and always is, except in lossless implementations (which I have never seen in JPEG, by the way).


Attached File(s) Thumbnail(s)
   
Find all posts by this user
Quote this message in a reply
Nov 5, 2013, 17:00 (This post was last modified: Nov 5, 2013 17:10 by foxmuldr.)
Post: #11
RE: Two questions...
(Nov 5, 2013 16:14)ofnuts Wrote:  
(Nov 5, 2013 13:41)foxmuldr Wrote:  This is not true. JPEG is lossy. It always loses color data. Anyone can see the difference through a simple test.

Of course... but it's not indefinitely lossy. You lose the data during the first few load/save cycles. After that, if you don't change the save parameters, the re-encoding of the decoded pixels eventually recreates the very same data, and there are no more losses.

I don't believe that's true, but I have not tested it. It is of interest to me and I will test it at some point with a C++ app I'll write using libpng and libjpg to compare successive images. Is there a way in GIMP to automatically save a history of the changes made to a document, such that the original image remains unchanged and the history of steps made to change the image are simply listed as a sequence of steps to reproduce the change at will?

I am thinking if you open each time and perform some edit, make the eyes more red, add some coloring to the feathers around the beak, alter the rust coloring on the nail, etc., basic edits, then you'll find that if you open it, make an edit, save it, open it, make another edit, save it, etc., that you'll lose far more color data. You could perform the same edits on a PNG file and do a compare from the last PNG to the last JPG and see the notable difference. It may not be a big deal to the human eye, but it's there, and those artifacts can creep up in applications as undesirable components, such as when being rendered in 3D as textures.

I think it's a moot point nowadays with the prominence of PNG. As data storage sizes increase, the need for saving more images on less space will become less, and lossless will become preferred over lossy because size is no longer an issue.
Find all posts by this user
Quote this message in a reply
Nov 5, 2013, 18:17
Post: #12
RE: Two questions...
(Nov 5, 2013 17:30)ofnuts Wrote:  
(Nov 5, 2013 17:00)foxmuldr Wrote:  I don't believe that's true, but I have not tested it.

Well, I have, see above(*). And you can do it too, no need for a C++ app. Just use the script above, or rewrite your own that does:

For JPEG number N:
- Take JPEG number N-1
- Add local edit, save as PNG number N
- Convert PNG number N to JPG number N

The intermediate PNG is used to convince the non-believers that ImageMagick cannot be using the original JPEG "encoding blocks" where it notices that the image hasn't changed.

(*) This is is also used for image forensics. The JPEG data "settles" with the save cycles, but additional edits unsettle them, so, if you receive a suspicious image, resave it with the very same parameters and compare the two images. There will be more differences in the recent edits. See http://fotoforensics.com/tutorial-ela.php (and see the caveat: "If an image is resaved multiple times, then it may be entirely at a minimum error level, where more resaves do not alter the image. In this case, the ELA will return a black image and no modifications can be identified using this algorithm. ")

I'll look into it.

Out of curiosity, why would someone open and save a JPEG image multiple times (or even more than once) without introducing additional edits each time?
Find all posts by this user
Quote this message in a reply
Nov 5, 2013, 23:30 (This post was last modified: Nov 5, 2013 23:35 by btraven.)
Post: #13
RE: Two questions...
I've done that--open then close an image without edits. A lot. I'll work on an image, close it and later think I can improve it just a little so I open it again, work on it without good result then close it without saving the edits.
Find all posts by this user
Quote this message in a reply
Nov 6, 2013, 11:04
Post: #14
RE: Two questions...
(Nov 5, 2013 23:50)ofnuts Wrote:  
(Nov 5, 2013 18:17)foxmuldr Wrote:  Out of curiosity, why would someone open and save a JPEG image multiple times (or even more than once) without introducing additional edits each time?

You always do some editing (otherwise you wouldn't save...) but the question is what happens to the areas of the image where you didn't make any change (as in my test images above).

I understand. I have several other projects ahead of my interest in this one. I have seen your script, but don't know enough about Python to be able to test it, even though I can run it. By using C++ and libjpeg I can know for sure what I am dealing with, and therefore verify it to myself.

My thoughts are two-fold:
(1) Of interest to learn about JPEG being not lossy after the Nth save.
(2) That it doesn't really matter any longer because of PNG's success, and low file size while maintaining its full losslessness. :-)

I plan to write a GIMP alternative at some point, but there is a large project ahead of that in my queue. When I get my large project out of the way, I'll come back to this as part of that alternative solution. Still haven't thought of a final name, though I am leaning toward a plugin module for an editor I'm writing, called WideEdit, the module to then simply be called Graphics. :-)

Thank you, ofnuts, for your replies.
Find all posts by this user
Quote this message in a reply
Nov 6, 2013, 11:06
Post: #15
RE: Two questions...
(Nov 5, 2013 23:30)btraven Wrote:  I've done that--open then close an image without edits. A lot. I'll work on an image, close it and later think I can improve it just a little so I open it again, work on it without good result then close it without saving the edits.

I mean to open a JPEG, and then save it without having made any edits. ofnuts was saying that after the 3rd save there is little new color information lost, as compared to the 1st save, or the 1st to 2nd, or 2nd to 3rd. His demo showed that after 10 iterations there was almost no loss, except along the edges, and around the white dots which were introduced at each edit.
Find all posts by this user
Quote this message in a reply
Post Reply 


Possibly Related Threads...
Thread: Author Replies: Views: Last Post
  A Few Questions anon_private 1 351 Sep 14, 2014 13:15
Last Post: rich2005
  Questions about GIMP forum 7 2,236 Mar 22, 2014 15:27
Last Post: rich2005
  Hi I'm new and I have some questions and suggestions. Clips 5 3,043 Jun 29, 2013 02:38
Last Post: Gamer_Z.
  Questions about Blur Tool dfro 2 1,116 May 31, 2013 14:59
Last Post: Grue
Question I have a few questions about 2.8 glathon 2 670 Sep 4, 2012 03:53
Last Post: glathon

Forum Jump:


GIMP ForumPortalArchiveContactTermsRSS