sRAW on 5D Mark II

[Alis]

Hello Everyone,

I had heard that using a lower than full resolution on the camera (going for sRAW at ~10 mpixel instead of full resolution RAW at ~20 mpixel) will improve the IQ due to camera not using interpolation to generate the full resolution image.

I tested this yesterday for the first time and it looked like it is pretty effective.

I am not concerned about file size at all. So, from an IQ standpoint only, what do the photography gurus of the Forum think? I am dying to know if this is overall worth it, if I am losing anything or I am safe doing this.

Cheers!

[herbert]

Hi Ali,

Using either sRaw or mRaw causes the camera to bin pixels together to reduce file size. This will result in a loss of information that cannot be recovered. If you want to maximise the image quality then you should shoot the full raw.

The advantage of s/m raw is that the files are smaller. This allows you to shoot more images on the same card and still maintain the post processing advantages of raw such as altering white balance and changing picture styles. It may be useful when you are getting close to running out of space on the card and you have no spares.

The smaller file size also allows the camera to make more use of the image buffer. So instead of being limited to 15 continuous shots in raw you may get more like 30 or 60. It will be a great advantage to action shots that require a few seconds of continuous shooting. One such example I have seen is a huge ski jump with 20+ images of the skier. With the difficult lighting of shooting on snow it is advantages to be able to shoot raw to correct white balance later and recover highlights.

Alex

[stuck]

I had heard that using a lower than full resolution on the camera (going for sRAW at ~10 mpixel instead of full resolution RAW at ~20 mpixel) will improve the IQ due to camera not using interpolation to generate the full resolution image.

That doesn't sound right at all. What you have heard implies that the 5D's sensor is only capable of capturing ~10mp (sRAW) and then it expands (interpolates) that up to ~20mp for a full res RAW file. I don't think so...

The 5D will capture ~20mp but if you opt to save sRAW then the camera will chuck away half of what it's captured and leave you with only ~10mp.

Ken

[Alis]

That doesn't sound right at all. What you have heard implies that the 5D's sensor is only capable of capturing ~10mp (sRAW) and then it expands (interpolates) that up to ~20mp for a full res RAW file. I don't think so...

The 5D will capture ~20mp but if you opt to save sRAW then the camera will chuck away half of what it's captured and leave you with only ~10mp.

Ken

Not really, Ken. The sensor does not see color. It sees the brightness. Over each one of the pixels it has a filter of one of the RBG colors and only one of them. So, interpolation comes into play when it tries to reconstruct or guess the two other components/channels of the actual picture.

What I mentioned I read what that with 22 mp available, the camera will do a better job at sort of guessing those other colors/channels. Not that the capture is at less than 22 and it fakes it (which in the sense I just mentioned, it kind of is ).

[Alis]

Hi Ali,

Using either sRaw or mRaw causes the camera to bin pixels together to reduce file size. This will result in a loss of information that cannot be recovered. If you want to maximise the image quality then you should shoot the full raw.

The advantage of s/m raw is that the files are smaller. This allows you to shoot more images on the same card and still maintain the post processing advantages of raw such as altering white balance and changing picture styles. It may be useful when you are getting close to running out of space on the card and you have no spares.

The smaller file size also allows the camera to make more use of the image buffer. So instead of being limited to 15 continuous shots in raw you may get more like 30 or 60. It will be a great advantage to action shots that require a few seconds of continuous shooting. One such example I have seen is a huge ski jump with 20+ images of the skier. With the difficult lighting of shooting on snow it is advantages to be able to shoot raw to correct white balance later and recover highlights.

Alex

Thanks, Alex. As I mentioned, I am not really in anyway in need of more space. So, file size is not a problem at all. When I go out, I carry more than a 100 GB cards with me and at home I use 2TB LaCie RAID drives to store than.

But I am really concerned about the quality. I really would love to know if this gives me a better quality, since I don't really need the huge resolution, I might as well just do sRAW for at least portrait.

Anyway, I was quoting Ken Rockwell:

"The smaller-sized images out of the 5D Mark II are spectacular. They are much, much sharper and cleaner than images from cameras on which that is their native resolution. When you start with over 20MP, it looks pretty good if you use all those to make 11MP or 5MP.

Why? Because they use less, or no, Bayer interpolation. No digital camera really resolves its rated resolution; they cheat and interpolate up, so at 100% at its rated resolution, no digital camera image is as sharp as a true scan from film.

At the 5MP setting, you have 100% R, G and B pixels, exactly as if you were using a Sigma Foveon sensor. If Sigma was selling this, they'd sell the 5MP (S) setting as if it were 15MP (also a lie).

What this means is that the lower resolution settings actually pack away lot more detail than you think. The S (5MP) setting of the 5D Mark II is a lot sharper than any 5MP camera."

Please let me know if you think differently now since I may have misquoted him.

Thanks!

[Photon Hacker]

Ali: For best technical image quality (IQ) always shoot ordinary raw. The required processing can be done in computer with total control if desired. Downsizing RAW files is processing, has possible logistical benefits of which the first is reduced storage requirements and sharpness is not one of them. I'm afraid that's just another fallacy from Ken Rockwell.

First, let's note that to compare sharpness images should be judged at the same scale. Comparing images of different sizes at 100% on screen (As Ken implies) is meaningless.

Sharpness is a subjective parameter of images. The main influence in sharpness is the quality of the detail at near and at the extreme of the resolution of our eyes. In other words, the detail at the finer level our eyes can see is what influences sharpness the most. Imaging systems (Lens and sensors) are better for coarser details than for finer details. Reducing the apparent size of an image naturally reduces the scale of the details, thus pushing the relatively blurry finer details in the image to the size where we can't see them and don't influence our perception of sharpness, instead we see coarser details in which the systems perform better. This is analogous as to the phenomenon that a halftone document looks ugly seen near, but seen at farther individual dots aren't perceived or can't be seen and we perceive the intended image rather than dots.

Full size RAW files can reproduce finer detail than downsampled RAWs, as you probably expected, and this will be noticed in a proper comparison. However, if they are both seen at 100% on screen you're comparing apples to oranges as explained above.

Proper downsampling will usually reduce noise, this can be explained easily by considering the implied low pass filtering which eliminates data at small scales, namely, some noise and some detail. The finite resolution of all optical systems including the human eye also acts as a low pass filter and also has this effect when a faithful print of a noisy photograph is seen small.

By the way, the Canon 5D is a different camera than the Canon 5D mark II and has different pixel count. Strictly speaking, neither of them and almost no DSLR captures pixels (Because only one color channel is registered at each photosite); sensels (Sensing elements) is a technically correct term.

Regards.

[Alis]

Ali: For best technical image quality (IQ) always shoot ordinary raw. The required processing can be done in computer with total control if desired. Downsizing RAW files is processing, has possible logistical benefits of which the first is reduced storage requirements and sharpness is not one of them. I'm afraid that's just another fallacy from Ken Rockwell.

First, let's note that to compare sharpness images should be judged at the same scale. Comparing images of different sizes at 100% on screen (As Ken implies) is meaningless.

Sharpness is a subjective parameter of images. The main influence in sharpness is the quality of the detail at near and at the extreme of the resolution of our eyes. In other words, the detail at the finer level our eyes can see is what influences sharpness the most. Imaging systems (Lens and sensors) are better for coarser details than for finer details. Reducing the apparent size of an image naturally reduces the scale of the details, thus pushing the relatively blurry finer details in the image to the size where we can't see them and don't influence our perception of sharpness, instead we see coarser details in which the systems perform better. This is analogous as to the phenomenon that a halftone document looks ugly seen near, but seen at farther individual dots aren't perceived or can't be seen and we perceive the intended image rather than dots.

Full size RAW files can reproduce finer detail than downsampled RAWs, as you probably expected, and this will be noticed in a proper comparison. However, if they are both seen at 100% on screen you're comparing apples to oranges as explained above.

Proper downsampling will usually reduce noise, this can be explained easily by considering the implied low pass filtering which eliminates data at small scales, namely, some noise and some detail. The finite resolution of all optical systems including the human eye also acts as a low pass filter and also has this effect when a faithful print of a noisy photograph is seen small.

By the way, the Canon 5D is a different camera than the Canon 5D mark II and has different pixel count. Strictly speaking, neither of them and almost no DSLR captures pixels (Because only one color channel is registered at each photosite); sensels (Sensing elements) is a technically correct term.

Regards.

Many thanks, Mario! Very good point.

BTW, I have 5D MKII not the older 5D.

[herbert]

Thanks, Alex. As I mentioned, I am not really in anyway in need of more space.

If you are not pushed for space on your card, or space in your buffer when continuous shooting, then do not shoot the smaller raw files.

As Mario described you maximise your image options, and so the ultimate quality, by capturing all the data.

Alex