Downsizing Canon 5Dsr raw files the smart way - how?

[Gel]

I recently purchased a Canon 5Dsr and it's very good but I don't always need the 50mp file size yet won't be shooting in the mRaw or sRaw sizes anytime soon and there is a difference in the file quality.

But would I be able to convert the Raw files to a smaller lossless DNG file, say 5000 pixels at the long edge instead. What would the resulting difference to the file be other than size, are there any extra benefits to this such as reduced noise and so on?

Update, appears I should of researched more, the only way to downsize the image is to use lossy compression anyway.

[biker]

But would I be able to convert the Raw files to a smaller lossless DNG file, say 5000 pixels at the long edge instead.

Only if you were able to persuade your camera to read just a part of your sensor. It wouldn't be RAW otherwise. I mean - RAW files are readings of physical red, blue, green and green pixels on your sensor. When downsampled, it's not a RAW.

Update, appears I should of researched more, the only way to downsize the image is to use lossy compression anyway.

You can still use a lossless compression in TIFF files (such as Deflate, Zip or LZW - but not JPEG). You can choose a 48bpp colour depth while developing from RAW to TIFF and downsample the TIFF as you wish. But it's not RAW anymore.

[nemophoto]

I recently purchased a Canon 5Dsr and it's very good but I don't always need the 50mp file size yet won't be shooting in the mRaw or sRaw sizes anytime soon and there is a difference in the file quality.

But would I be able to convert the Raw files to a smaller lossless DNG file, say 5000 pixels at the long edge instead. What would the resulting difference to the file be other than size, are there any extra benefits to this such as reduced noise and so on?

Update, appears I should of researched more, the only way to downsize the image is to use lossy compression anyway.

I have never seen a qualitative difference when shooting mRaw or sRaw from my 5Ds (or any of the recent cameras). Is it a file storage issue you're dealing with and, hence why you don't want the full RAW? I've used the mRaw on my 5Ds quite a bit and never had any issues (still loads of detail, no difference in noise, et al). The only caveat is that not all converters can deal with the mRaw (notably, if you use it, DxO). Otherwise, most (such as Lightroom) do a good job.

[BobShaw]

Mate, if you don't want the raw file size then don't buy a 50MP camera. Even if you don't use the capability now, you may in the future. One day you may go back to that Pulitzer Prize winning shot and find that you had downsized it. I know I have gone back to use images I shot in the early digital days and been disappointed that the first few cards were done in JPG. Storage is one of the cheaper things in photography.

[Tony Jay]

I have to agree with Bob here - why get a camera with such a large sensor if you don't want what it provides?
What is it that we don't understand?

Tony Jay

[Craig Lamson]

I have to agree with Bob here - why get a camera with such a large sensor if you don't want what it provides?
What is it that we don't understand?

Tony Jay

Then let me give you my reason.
On the projects I shoot, some images are used large by the client and others will never be used larger than say 5x7.  As such I deliver the major images at or near full Rez and the others at 12 inches at the long side. 

So I either downsize (I tried s and m raw and prefer downsizing the finished file instead) or I shoot with two cameras ( and still downsize).  Two cameras makes no sense.

[nputtick]

Then let me give you my reason.
On the projects I shoot, some images are used large by the client and others will never be used larger than say 5x7.  As such I deliver the major images at or near full Rez and the others at 12 inches at the long side. 

When you shoot them, I assume you don't know which will end up as large and which as small, so they all need to start off large in the camera, i.e. full sized RAW files.  Then I would batch process all of them to DNG, resizing at the same time to 5100 x 3400 (this is nearest to your request for 5000px long side), and have a coffee while the files are processing.  There are many tutorials how to do this.  Here's one that's pretty easy to follow:

https://www.youtube.com/watch?v=h4YB3KAkkK4

Then you would probably want to convert them all to jpegs too, so they can easily be viewed on the client's computer.  Once the client has chosen which image they want large, and which small, you can go back and do your final processing - large images from the original RAW, 7x5 jpegs from the smaller files.

I don't have any issues with you buying a camera with a high res sensor and wanting some as smaller files.  You have the best of all worlds with that camera IMO.

Cheers
NIgel 

[Petrus]

Storage is so cheap now, that I see no reason to make small RAWs out of big ones, even if it was possible (DNG files).

In Lightroom you can rate (stars) or code (colors) each usable original big RAW file individually, then export in any size you want (resize option in export dialog). I fail to see why you would want small RAW files for the pictures which end up small. Too complicated, with the cheap storage we now have.

[Craig Lamson]

When you shoot them, I assume you don't know which will end up as large and which as small, so they all need to start off large in the camera, i.e. full sized RAW files.  Then I would batch process all of them to DNG, resizing at the same time to 5100 x 3400 (this is nearest to your request for 5000px long side), and have a coffee while the files are processing.  There are many tutorials how to do this.  Here's one that's pretty easy to follow:

https://www.youtube.com/watch?v=h4YB3KAkkK4

Then you would probably want to convert them all to jpegs too, so they can easily be viewed on the client's computer.  Once the client has chosen which image they want large, and which small, you can go back and do your final processing - large images from the original RAW, 7x5 jpegs from the smaller files.

I don't have any issues with you buying a camera with a high res sensor and wanting some as smaller files.  You have the best of all worlds with that camera IMO.

Cheers
NIgel

Nope,  know when I shoot which filed will be downsized. The extra step of creating dng's really makes no sense to me.

[xpatUSA]

I recently purchased a Canon 5Dsr and it's very good but I don't always need the 50mp file size yet won't be shooting in the mRaw or sRaw sizes anytime soon and there is a difference in the file quality.

I am not a Canon owner but I've been thinking lately about resolution and viewing stuff with an emphasis on image quality. Out of interest, may I ask a question or two?

By "file quality" may I assume that you mean "image quality"?

With the smaller sizes, say sRaw, how do you determine quality?

I ask because my own research and tests say that if you view whole images on a monitor 'fitted to screen' it's hard to tell any difference in quality; but if you zoom in to 100% or more, differences become more visible but that would be unfair to the sRaw image (pixelation, etc.).

I suppose I'm hinting that, with a different point of view, sRaw files might become acceptable . . .

Off topic a bit, but I habitually shoot my Sigma SD1 Merrill in low-resolution raw 2236x1568px because a) the image quality is a bit higher and b) I only view my images on my monitor (I don't print anything).

[BobShaw]

Having just had a play with a 5Ds myself, I would say this. Why don't you just shoot one card large raw and one card small raw and decide later on based on the requirement. There is no such thing as lossless DNG.

[xpatUSA]

There is no such thing as lossless DNG.

I've read that the DNG spec was not "lossy" until 2012 when they introduced "lossy compression" in V 1.4

So, are we talking about other forms of loss?

[BobShaw]

So, are we talking about other forms of loss?

DNG does not carry over all over the proprietary camera information, which is why most of the camera manufacturers don't use it. If you bring out a new feature and DNG doesn't support it then it limits the ability to create new features. That is why Hasselblad dropped DNG support only 2 years after doing it. There used to be a much better explanation than mine on the Hasselblad site but it seems to have disappeared.

[Denis de Gannes]

DNG does not carry over all over the proprietary camera information, which is why most of the camera manufacturers don't use it. If you bring out a new feature and DNG doesn't support it then it limits the ability to create new features. That is why Hasselblad dropped DNG support only 2 years after doing it. There used to be a much better explanation than mine on the Hasselblad site but it seems to have disappeared.

The camera manufacturers will never support DNG and allow Adobe to dictate their ability to be creative, why? Adobe is not even a camera manufacturer.

[xpatUSA]

DNG does not carry over all over the proprietary camera information, which is why most of the camera manufacturers don't use it. If you bring out a new feature and DNG doesn't support it then it limits the ability to create new features. That is why Hasselblad dropped DNG support only 2 years after doing it. There used to be a much better explanation than mine on the Hasselblad site but it seems to have disappeared.

Quite so and there's more. My (Sigma) cameras have a separate 3x3 matrix for each and every white balance. All DNG 1.4 offers is Adobe's .dcp-style of incandescent WB and 65K WB interpolated for anything in-between and only a single multiplier for each channel at that. Although there is an 'as shot' WB tag, it too has only a single multiplier for each channel, not Sigma's 3x3 matrix. In other words, at least for Sigmas, most of the DNG tags are compromises and represent someone's best guesses - as opposed to the very large number of scene- and sensor-specific tags in an actual X3F file.

It would not surprise me if other camera makes have similar difficulties in correct rendition - even though the raw image data itself is not changed! Er, by "correct rendition, I mean as the the camera manufacturer intended - not as Adobe thinks it should be rendered . . .

[scyth]

All DNG 1.4 offers is Adobe's .dcp-style of incandescent WB and 65K WB interpolated for anything in-between and only a single multiplier for each channel at that.

it is technically not a single mulitplier per channel (Adobe operates on demosaicked data, not on raw channels) and then you can supply WB for each illuminant yourself with your own profiles - no need to use standard Adobe profiles...

[xpatUSA]

it is technically not a single mulitplier per channel (Adobe operates on demosaicked data, not on raw channels) and then you can supply WB for each illuminant yourself with your own profiles - no need to use standard Adobe profiles...

Thank for the clarification. Yes, by "channel" I did mean R,G,B channels and by one per channel I meant effectively a diagonal matrix.

So by "each illuminant" that would be a separate .dcp for each illuminant?

Pardon my ignorance, I don't use Adobe at all but I do remain interested in how DNG works, nevertheless.

[scyth]

Thank for the clarification. Yes, by "channel" I did mean R,G,B channels and by one per channel I meant effectively a diagonal matrix.

unlike subtractive math mess in Foveon you are quite fine with a single multiplier for a true raw ("RGB" model) channel... in any case a proper implementation shall write illumination spectrum + CFA transmission spectral data + optics spectral transmission + whatever else required...

So by "each illuminant" that would be a separate .dcp for each illuminant?

and in DNG container you can store multiple dcp profiles inside... so you can have a raw file written by a camera firmware in DNG format with as many profiles tuned for as many illuminants as you want...