RAW vs. cRAW

Fuzzphoto wrote: No two images are the same, and as long as you need to make two separate exposures to show the difference between the two formats, you aren't proving anything. So, it's simply technically impossible to provide conclusive proof, unless you have access to the camera's firmware and are able to derive both file formats from the same exposure. I'm convinced Sony provided both file formats in the first place to avoid uncompressed-RAW proponents switching to competing brands.

However you could create a studio set up with extreme lighting (fixed not flash) and color conditions and contrast where you could shoot 2 frames remote controlled same SS and F and the only difference would be the format. After that you could pixel peep those areas where peopel claim there woudl be image loss. Keep in mind we are not looking for data differences we are looking for image differences detectable by the eye. I suspect you could see no difference.

One interesting difference; a full raw converts to a smaller DNG file size than does a cRAW file.

(from a850)
full size raw - 36,501 KB
DNG conversion - 15,102 KB

cRAW           - 24,483 KB
DNG conversion - 19,921 KB

Adobe's LR3/PS CS5 has won the war (for the moment) as far as raw converters go. Should another superior converter appear, I think DNG has a good chance of being supported.

I'm comfortable with DNG as a long term file format. As far as I know, Bibble is the lone holdout that doesn't work with converted DNG files (only in camera DNG's).

Graham

It is really interesting. The only guess is LR tries to approximate or predict lost bits. But even so it is intreting why DNG from cRAW is bigger.
But one question - is it correct that both RAW and cRAW contained the same (as it possible to make) scene? Coz different level of small objects with high contrast will change DNG significantly, coz Huffman compressed files depend on content, but RAW and cRAW not (size is different due to embedded preview jpg).

Oh, not a again

Short version: images in cRaw are for all intent and purposes equal to Raw images. I can confidently predict based on the technical format, that no-one will ever produce a 'realistic' photo where the difference can be noticed, and I don't think anyone can even produce an 'artificial' photo to show the difference.

As a consequence, cRaw is clearly preferable to Raw since the files are much smaller. FPS is potentially higher than in Raw since the data-pipeline is the bottleneck, the compression on cRaw is extremely fast. Note, that after expansion, we have 12-bit Raw file again, so in lightroom and photoshop you work in the full color space.

I believe that the only reason we also have the Raw format is to prevent bad press from 'not having full raw' fanatics.

----
cRaw divides the image in 16 pixel groups, and stores the maximum and mininum pixel in each group as 11-bits. This is a 'loss' of 1 bit but since it is on the maximum and minimum, this is below the normal noise of the sensor, bayer array, etc. The other 14 pixels are stored as a 7 bit number that gives the relative color with respect to the minimum and maximum value. If the difference between the minimum and maximum is lower than 7 stops, than this is exact (no loss). If it is higher, there is a potential loss of precision of up to 4 bits. (not 5 since we can round). However, in a such a high-difference range (very bright to very dark) there is already a big loss of fidelity on the sensor level, bayer filter, etc. So, even that loss is technically within the 'noise' level anyway.

All in all, cRaw is a very smart and efficient (nearly loss-less) encoding of Raw files. For almost all images, there is no difference at all with the Raw file. And based on the technical algorithm, I agree that there will be perhaps 1 in a 10.000 images where a difference might perhaps be noticable -- perhaps

---
btw. I think cRaw is also preferable to the DNG format since DNG files do not always preserve the 'makernote' correctly. In particular, the recent Sony lens information block is not preserved. See more info about this in this thread post. Also, if you edit a DNG in lightroom or camera raw, the XMP edit information is stored in the DNG itself instead of as a seperate side-car XMP file which is not always preferable from a backup perspective, and from a safety perspective (ie. disk write failures).

Edited by Vivec - 16 August 2010 at 20:57

Vivec wrote: Oh, not a again Short version: images in cRaw are for all intent and purposes equal to Raw images. I can confidently predict based on the technical format, that no-one will ever produce a 'realistic' photo where the difference can be noticed, and I don't think anyone can even produce an 'artificial' photo to show the difference. As a consequence, cRaw is clearly preferable to Raw since the files are much smaller. FPS is potentially higher than in Raw since the data-pipeline is the bottleneck, the compression on cRaw is extremely fast. Note, that after expansion, we have 12-bit Raw file again, so in lightroom and photoshop you work in the full color space. I believe that the only reason we also have the Raw format is to prevent bad press from 'not having full raw' fanatics. ---- cRaw divides the image in 16 pixel groups, and stores the maximum and mininum pixel in each group as 11-bits. This is a 'loss' of 1 bit but since it is on the maximum and minimum, this is below the normal noise of the sensor, bayer array, etc. The other 14 pixels are stored as a 7 bit number that gives the relative color with respect to the minimum and maximum value. If the difference between the minimum and maximum is lower than 7 stops, than this is exact (no loss). If it is higher, there is a potential loss of precision of up to 4 bits. (not 5 since we can round). However, in a such a high-difference range (very bright to very dark) there is already a big loss of fidelity on the sensor level, bayer filter, etc. So, even that loss is technically within the 'noise' level anyway. All in all, cRaw is a very smart and efficient (nearly loss-less) encoding of Raw files. For almost all images, there is no difference at all with the Raw file. And based on the technical algorithm, I agree that there will be perhaps 1 in a 10.000 images where a difference might perhaps be noticable -- perhaps --- btw. I think cRaw is also preferable to the DNG format since DNG files do not always preserve the 'makernote' correctly. In particular, the recent Sony lens information block is not preserved. See more info about this in this thread post

Agree, not again, but happened. :)

I do not agree that 11 bit in best case is the same as 12, I do not agree that there is no differences especially on 24MP printed in big format, but at the same time I do believe that it is almost impossible to see differences. And I do not agree that cRAW is preferable over DNG or RAW. If somebody has a lot of storage then RAW is preferable. If storage is expensive then may be even jpg would be preferable for some fotographers. Regarding DNG I already mentioned that it is not perfect, but for me it is very acceptable. But any way I do agree that better to stop this discussion :)

I do remember the mp3 inovation time. At that moment a lot of people told that there is no differences and they loafing from others who told that they could hear the difference. The same I can say about Dolby True HD and lossy Dolby Digital. A bit wrong comparation between sound and image but I hope it helps to show my opinion.

Vivec wrote: Oh, not a again

+1

Apart from gnatsum's spectacular contribution this is all old news.

gnatsum wrote: Raw is 12 bit and CRAW is 8 bit. You lose all the highlight recovery capabilities that 12 bit raw files offer. If you shoot in very controlled environments and know there is no chance of some highlights blowing out, then shoot CRAW if you wish. I tried shooting CRAW for a few months this year, and really it was not worth saving space for losing pictures to blown out highlights...

Please gnatsum, give us some evidence.

gnatsum wrote: Raw is 12 bit and CRAW is 8 bit. You lose all the highlight recovery capabilities that 12 bit raw files offer. If you shoot in very controlled environments and know there is no chance of some highlights blowing out, then shoot CRAW if you wish. I tried shooting CRAW for a few months this year, and really it was not worth saving space for losing pictures to blown out highlights...

This is totally wrong... In best case it is 11 bits. Also it is totally wrong that highlights will be blown up (at least by cRAW format). It is no matter highlight midrange or lowlight - important that 16 pixels block will have values which won't fit cRAW range.
But unfortunately looks like only several people tried to understand format. Most try to ague about without any checks of format or even making comparations.

PS. For those who still have time I have very nice new topic: Is JPG with 100% quality settings has any visible differences as JPG with 90%