How much practical image quality is really lost with RAW compression on the D200? As I now understand it, the D70’s I’ve been using have always had RAW compression (no choice) and I have found the image quality on those to be excellent. Now I have a choice and don’t know if the increase (slight, lots, none?) in image quality is worth the increased file size (2x!). If the compressed RAW quality is just as good for any practical purpose (like prints up to 11x14 or stock photography sales) as the full RAW file then I can save some money that I would otherwise have to spend on both camera and computer storage space.
I’m interested in real world user observations or tests concerning this, not so much what someone read or what some techie guru suspects. I’m also interested in knowing how I can effectively test this for myself. What subject matter and image parameters would give me the best “look” at any appreciable image quality loss.
Any real help appreciated. I have about a week to figure this out before our trip.
I shoot uncompressed. I can get about 120 pics on a 2 gig card. I really don't know that it makes a difference, except that it fulfils my urge to start as clean and artifact free as possible. I guess it depends on how much you want to print, as opposed to post, and how much you care to store on your HD before you get preventative. What I try to do is weed, weed, weed the bad and/or non-useable ones out. I regularly file originals on DVDs, which really takes the pressure off.
Phil Hodgkins
Savannah, GA
Old Geezer having fun
Posted by Richard Ditch on July 27, 2006 at 11:44 AM
Like Phil I just decided to shoot uncompressed raw when I upgraded from the D70 to the D200. Why worry about what you might be giving up with compressed files. CF card prices have really dropped, and I added a 4GB and 2 2GB CF cards to my stock when I added the D200. With the older cards from the D70 I now have 13.5 GB of CF storage when I'm out shooting.
I've read an on-line article that explains how Nikon does the compression, and the conclusion was that any degradation would show up as lost highlight detail. So, if you wanted to see if it matters to you set up a test that emphasizes the highlights.
Posted by Mike Maples on July 27, 2006 at 12:11 PM
I did just that with my stuffed panda yesterday. There is no difference that I can see at 100% magnification with RSP on a 21" calibrated monitor. In fact, I had some blown highlights that were actually slightly (very, very slightly) worse on the uncompressed image! Go figure. The histogram showed a very, very tiny shift to the left for the compressed image but it was probably because there were two shot taken and they are never perfectly identical. It made no difference in the image as viewed.
Detail, color and everything else (except file size...8.2MB vs 15.3MB) looked identical. My D200 is now officially set to compressed RAW!
Posted by Richard Ditch on July 27, 2006 at 12:48 PM
I interpret your test results as showing that the uncompressed file captured a wider range of tones - hence the indication of more blown highlights and the shifted histogram for the compressed file.
You've got more data in the uncompressed capture, so you'll need to tweak the settings in your raw converter to use as much of that data as possible.
That's very interesting Mike. Everything I've read concernng the RAW compression has said much the same thing, but I've never actually heard any kind of percentages of difference in a controlled test. I don't know if this exists, but would love to see it if it does. 8.2 vs. 15.3 is a BIG difference. I don't know what the difference is on the D2X is but I assume it's similar. Like I say, VERY interesting. I agree with Paul that you will always lose something, but there's a large difference between the theoretical and the real world. If I can't see a difference when it's printed at 16 x 20, etc. then there's no difference in my opinion. Has anyone out there run any tests with actual prints?? Anyone? Bueller? Bueller?
I get similar results to Mike when comparing file size of compressed vs. uncompressed raw files. Some compressed raw files go as high as 10 MB depending on the detail in the image. One thing to consider is how fast you need your data written to your CF card. The smaller compressed files will take less time to write to the card. If you are shooting sports or other photojournalistic work, your priority may be fast write time so as to clear the camera's buffer and capture as many shots in as short a time as possible. If you are shooting still life or studio or similar work that isn't time bound or moving, uncompressed will give you the most data, whether the difference is or is not noticable. I think you would have to understand the compression algorithm in order to design the right kind of test scene to shoot in order to quantify how much the compressed raw file differs from one that is uncompressed.
One thought I had was to shoot the exact same controlled scene with the exact same settings, 2 uncompressed images and 2 compressed images. Use the exact same raw conversion parameters for all four images and do image differences. Differences between the two compressed images would give you an idea of what variance the raw conversion process is introducing. The same goes for differences between the two uncompressed images. Then do a difference between the compressed and uncompressed and somehow factor in the like-image variances. This would be a complicated test. I don't know if it makes scientific sense, but it seemed to make sense to me conceptually.
Any thoughts on this test procedure? Would it produce the desired knowledge? What are the faults with the proposed test?
- - - I had once thought about doing a few print tests, but then I couldn't find any good reason to do one.
What's the point? If a pixel is compressed then it is "altered" and there is nothing theoretical about it.
If you crate images good enough to print to 16x20, then they could be good enough for major uses, so why the hell would you want just "good quality" images when you could have top quality ones? Your compressed image may work great for simple prints, but they will not be good enough for any high paying major uses,
so why would you even want to compress in the first place?
Just to save card space? Buy more memory cards!
It's not like anyone here is shooting 20K frames for Nat'l Geographic and needs the space.
Discussion
8 repliesHow much practical image quality is really lost with RAW compression on the D200? As I now understand it, the D70’s I’ve been using have always had RAW compression (no choice) and I have found the image quality on those to be excellent. Now I have a choice and don’t know if the increase (slight, lots, none?) in image quality is worth the increased file size (2x!). If the compressed RAW quality is just as good for any practical purpose (like prints up to 11x14 or stock photography sales) as the full RAW file then I can save some money that I would otherwise have to spend on both camera and computer storage space.
I’m interested in real world user observations or tests concerning this, not so much what someone read or what some techie guru suspects. I’m also interested in knowing how I can effectively test this for myself. What subject matter and image parameters would give me the best “look” at any appreciable image quality loss.
Any real help appreciated. I have about a week to figure this out before our trip.
I shoot uncompressed. I can get about 120 pics on a 2 gig card. I really don't know that it makes a difference, except that it fulfils my urge to start as clean and artifact free as possible. I guess it depends on how much you want to print, as opposed to post, and how much you care to store on your HD before you get preventative. What I try to do is weed, weed, weed the bad and/or non-useable ones out. I regularly file originals on DVDs, which really takes the pressure off.
Phil Hodgkins
Savannah, GA
Old Geezer having fun
I've read an on-line article that explains how Nikon does the compression, and the conclusion was that any degradation would show up as lost highlight detail. So, if you wanted to see if it matters to you set up a test that emphasizes the highlights.
Detail, color and everything else (except file size...8.2MB vs 15.3MB) looked identical. My D200 is now officially set to compressed RAW!
You've got more data in the uncompressed capture, so you'll need to tweak the settings in your raw converter to use as much of that data as possible.
Compress any file and you ALWAYS lose something.
The rest is just simple logic depending on what you want in the end.
One thought I had was to shoot the exact same controlled scene with the exact same settings, 2 uncompressed images and 2 compressed images. Use the exact same raw conversion parameters for all four images and do image differences. Differences between the two compressed images would give you an idea of what variance the raw conversion process is introducing. The same goes for differences between the two uncompressed images. Then do a difference between the compressed and uncompressed and somehow factor in the like-image variances. This would be a complicated test. I don't know if it makes scientific sense, but it seemed to make sense to me conceptually.
Any thoughts on this test procedure? Would it produce the desired knowledge? What are the faults with the proposed test?
What's the point? If a pixel is compressed then it is "altered" and there is nothing theoretical about it.
If you crate images good enough to print to 16x20, then they could be good enough for major uses, so why the hell would you want just "good quality" images when you could have top quality ones? Your compressed image may work great for simple prints, but they will not be good enough for any high paying major uses,
so why would you even want to compress in the first place?
Just to save card space? Buy more memory cards!
It's not like anyone here is shooting 20K frames for Nat'l Geographic and needs the space.