Is reducing jpg size inorder to save space a good idea?

[jerseyboy]

I work together with a colleague maintaining an archive of our photos from many different sources and in all shapes and sizes.

He wants to reduce the size (not the physical dimensions of the image but the space that it occupies on the disk) of all the jpg's over 3MB (some are over 8MB) down to 3MB in most cases by resaving the image in Photoshop from Quality 12 or 10 down to 8.

He says that there is no visible difference on the screen in the image once saved in a reduced size and it makes for quicker transfers and reduced storage space.

I have a gut feeling that this is a bad move but am incapable to produce any good arguments in my favour. Can anyone help me?

Thanks.

 

[Tom Mann]

Tens of thousands of pages (if not more) have been written on the best practices for archiving images. Just Google {best practices archiving digital photos} and read a few of the articles.

Here's the 1 minute summary:

a) Don't use JPGs to archive images because the flip of just one bit almost anywhere in the JPG file (aka, "bit rot" - Google the term) can bizarrely corrupt the remainder of a JPG image, rendering it essentially useless without taking extremely time-consuming measures to fix the file.

b) If you insist on storing your image archive on JPGs, the appropriate degree of compression (aka, quality factor) should be determined by the anticipated future use of these images. If you anticipate further work on an image, only use very high quality factors (say, 10 or higher), but if you are sure that no future work on an image will be needed and the anticipated audience is not very discriminating (eg, no prints, just small sized display on a monitor), you can probably get by with quality factors as low as 7.

HTH,

Tom M

 

[jerseyboy]

Many thanks for your very informative response and thanks to it I have found an excellent site dealing with the subject of archiving photos and all it's ramifications http://archivehistory.jeksite.org

However, I was rather hoping to find a few simple and clear arguments to show to a person who sees an image which is 6000 x 4000 pixels saved in JPG "12" in Photoshop and which "weighs" 8MB and who would rather resave it again in Photoshop in "8" "at 6000 x 4000 pixels "weighing" this time 3 MB.

He says that on the screen there is no difference so why have a file of 8MB when it can be 3MB?

 

[TonyCooper]

That "no difference" argument related to ppi. An image displayed on a computer screen at 72 ppi is just as clear
as an image displayed at 300 ppi. But, if you want to print that image, the 300 ppi is what should be used.

There are arguments that the 72 ppi figure is too low, and that it should be 120 ppi, but 72 seems fine to me.

Personally, I shoot RAW and keep the RAW files on external drives, and save my .jpgs at a crop at 72 ppi
(usually 1800 x 1200) for web display. When I want to print an image, I go to the RAW file.

 

[thebestcpu]

Hi Jerseyboy
The references provide a lot of good information. It also depends on usage.
Taking your comments as is. I took a 4000x6000 pixel image, created a slowly changing low noise color gradient across the field (such as you might see with the sky), and them saved the image a) as is with no compression PSD b) JPEG 12 Quality c) JPEG 8 Quality and d) JPEG 4 quality.
Those are shown in the image below.

I used the gradient because already being in 8 bit mode you can be prone to banding and it gets worse with compression.

The image below also depends on the color space used (I used sRGB and expect wider gamut color spaces to have worse results, your monitor, your profile for you monitor etc etc

The tip image is the image at 12.5% magnification and the image below is the same magnification yet only displaying the Luminosity component.

To me, there is banding in all and JPEG 12 already shows a banding degradation. It is somewhat worse at JPEG 8, and definitely worse at JPEG 4

Not sure if this helps or not yet thought an empirical example over a theoretical or worst case academic approach might be helpful.

 

[Tom Mann]

That "no difference" argument related to ppi. An image displayed on a computer screen at 72 ppi is just as clear
as an image displayed at 300 ppi....

This statement can be made even stronger: The ppi number that one can set in some image files makes absolutely no difference when the image is displayed on-screen because it is completely disregarded in this case. Each pixel in the image is mapped to a pixel on screen, completely independent of this number. In fact, for many methods of printing (eg, printing kiosks, drug-store level printing services, and even many on-line printing services) any ppi number specified in the image file also isn't used. You send them an image and tell them what size (in inches or cm) you want your image to be, and they will re-sample it to whatever ppi is required.

 

[Tom Mann]

Many thanks for your very informative response and thanks to it I have found an excellent site dealing with the subject of archiving photos and all it's ramifications http://archivehistory.jeksite.org

However, I was rather hoping to find a few simple and clear arguments to show to a person who sees an image which is 6000 x 4000 pixels saved in JPG "12" in Photoshop and which "weighs" 8MB and who would rather resave it again in Photoshop in "8" "at 6000 x 4000 pixels "weighing" this time 3 MB.

He says that on the screen there is no difference so why have a file of 8MB when it can be 3MB?

As I said in my previous post, it all depends on the anticipated future use. If all you anticipate ever using the image for is some sort of "reminder" of what the image contained, or if you anticipate that you will never have to do any further processing to the image, sure, compress it. However, if you anticipate using this image at some point in the future to produce other variants, you want as much quality as possible. The reason is that with each generation, the loss of quality keeps adding up, and after a few JPG saves at mid-quality levels, it will start to look really rough. This is why they call it a "quality" setting.

Tom M

 

[jerseyboy]

Hi Jerseyboy
The references provide a lot of good information. It also depends on usage.
Taking your comments as is. I took a 4000x6000 pixel image, created a slowly changing low noise color gradient across the field (such as you might see with the sky), and them saved the image a) as is with no compression PSD b) JPEG 12 Quality c) JPEG 8 Quality and d) JPEG 4 quality.
Those are shown in the image below.

I used the gradient because already being in 8 bit mode you can be prone to banding and it gets worse with compression.

The image below also depends on the color space used (I used sRGB and expect wider gamut color spaces to have worse results, your monitor, your profile for you monitor etc etc

The tip image is the image at 12.5% magnification and the image below is the same magnification yet only displaying the Luminosity component.

To me, there is banding in all and JPEG 12 already shows a banding degradation. It is somewhat worse at JPEG 8, and definitely worse at JPEG 4

Not sure if this helps or not yet thought an empirical example over a theoretical or worst case academic approach might be helpful.

Yes thanks, very helpful but I am sure that my colleague will say that the difference between 12 and 8 is "negligible".

Is there and example that I could uses that could "exaggerate" the differences, without falsifying them of course?

 

[Tom Mann]

OK... see if the attached example is sufficiently convincing.

Attached is a photo of a building that I took through the tinted windshield of my car on a gray overcast day. Having seeing the building in bright sunlight, I knew it had much more interesting colors and shadowing, so I attempted to bring this out in the shot I took under the bad lighting conditions by fairly agressive post processing. This involved increasing the local contrast, vibrance, saturation, and a final color correction step. This is the 1st frame of the attached GIF animation.

Next, for the purpose of comparison, I re-saved the original JPG at a quality = 7 setting and performed *exactly* the same post processing steps on this "slightly" degraded version of the image. A comparison of the two results is shown in the attached GIF animation. Note: the in-forum preview of the GIF won't show the animation. Important note: You have to right click on the in-forum preview and to open the full resolution version in its own window or tab, and then wait for the full file to load and start to animate. Also, FYI, the original was about 4000 pixels horizontally, but I cropped it down to around 1200 px so the forum uploading software would accept it and it wouldn't take even longer to download the entire file.

If the above demo doesn't convince your colleague, there are plenty of similar demos available on the web, the major difference being that for most of these, multiple re-saves at moderate quality factors were performed, versus the single re-save (but substantial post processing) in my demo. Here are some links to other demos of the loss of quality:

http://masteryournikon.com/2013/01/18/dont-save-a-jpeg-file-twice/

http://www.google.com/imgres?imgurl...QVcKrGInp-QHp7YbICw&tbm=isch&ved=0CFAQMygpMCk
(10 re-saves at Q=90% as described here: http://www.dpreview.com/forums/post/33551084_

http://www.google.com/imgres?imgurl...QVcKrGInp-QHp7YbICw&tbm=isch&ved=0CHgQMyhRMFE (various numbers of re-saves)

http://www.coxrail.com/scanning/Scan-JPG-multi-saves.htm (10 re-saves, but he compared different programs)

General articles of JPG compression artifacts:

https://en.wikipedia.org/wiki/Compression_artifact

http://www.webdesignerdepot.com/2010/03/techniques-for-squeezing-images-for-all-they’re-worth/


HTH,

Tom M

PS - Let me reiterate a couple of comments I made twice before, saying it a bit differently this time: The optimal trade-off between compression induced artifacts and saving money simply can not be made in any intelligent way if your organization can't say how the saved images are likely to be used. If they are not likely to be further processed, you probably can accept a Q of 7 or 8, but if they are likely to be post processed and quality is important I would never intentionally corrupt my images by this amount. Also, if you really want to drive your colleague up the wall, have him read the following articles on the effect of inevitable "bit rot" on JPGs, noting that the only ways around this is to store your images in uncompressed form or use more advanced file systems:
http://arstechnica.com/information-...-and-atomic-cows-inside-next-gen-filesystems/
http://www.techtimes.com/articles/3...t-and-digital-dark-age-but-dont-panic-yet.htm
http://www.slideshare.net/hakbailey/whs-talk-201404web (great little summary of good practices)

Examples of the effects of bit rot on JPGs:
http://www.google.com/imgres?imgurl...RVbHeLonj-QHemKOAAw&tbm=isch&ved=0CC4QMygRMBE

http://www.google.com/imgres?imgurl...RVbHeLonj-QHemKOAAw&tbm=isch&ved=0CEcQMyggMCA

http://www.google.com/imgres?imgurl...RVbHeLonj-QHemKOAAw&tbm=isch&ved=0CEoQMygjMCM

http://www.google.com/imgres?imgurl...RD4Kw-AHxqLD4DQ&tbm=isch&ved=0CCIQMygeMB44yAE

 

[jerseyboy]

OK... see if the attached example is sufficiently convincing.

Grateful thanks for the time and the trouble that you have spent compiling the message above.

Extremely useful links and your attached image is most convincing.

 

[Tom Mann]

You are quite welcome! Let us know if we can help you in any other way.

Best regards,

Tom M