Sofware for down-sizing avoiding demosaicing ?

[Francois]

I think there's a misunderstanding somewhere...

I do not see why you would get a better image from using a 'binned' image (of 1/4 the full size),
instead of working with the full sized demosaiced image.

I tried to explain my reasons, but those could very well be wrong or not clearly explained.

With the binning in camera, there is a gain in ISO performance, like they do for astronomic photography, or in the 41MP Nokia phone.

You don't get this ISO boost (or all of it) if you do the binning in PP, it has to be before a certain step in the image pipeline. I'm sorry, I didn't get the detailed information, and couldn't explain it here. But I read it from several different sources about pixel binning.

[xpatUSA]

Intersting article, thanks.

I also read that to be effective, the binning has to be done in-camera, at the sensor level. If it's done in PP, it combines read noise, and it is less effective.

Reading a little bit on the subject also makes me understand why the method I was looking for is wrong : avoiding demosaicing, I would maybe have a slight gain in sharpness, but none in noise. I would even be much worse, as I would have lost so much information.

For a gain in high iso, I could see how the binning is done in-camera with the D800. It looks that maybe John was right, depending on how Nikon produces smaller JPG files. But I'd lose the high bit rate of raw... Well, maybe only field-testing will answer this.

Bonjour Francois,

I have two Foveon cameras and, in pursuit of sharp watch images, I used their half-size true sensor pixel binning (VPN) for a long time. However, it turned out in the end that the simple averaging performed on-sensor is actually inferior to the use of modern down-sampling algorithms use by Adobe, et al, both in terms of noise reduction and sharpness (e.g. Lanczos). This statement applies more to images for viewing on a monitor than for huge prints.

Does the D800 really do true binning, by the way?

[Francois]

Bonjour Francois,

I have two Foveon cameras and, in pursuit of sharp watch images, I used their half-size true sensor pixel binning (VPN) for a long time. However, it turned out in the end that the simple averaging performed on-sensor is actually inferior to the use of modern down-sampling algorithms use by Adobe, et al, both in terms of noise reduction and sharpness (e.g. Lanczos). This statement applies more to images for viewing on a monitor than for huge prints.

Does the D800 really do true binning, by the way?

Hi Ted,

Thank you for this useful information. Even if it says the opposite of what I've read in theory, I guess sometimes you really have to try.

About the true binning of the D800, I don't know, this is still someting I have to read about. But I'm not v. optimistic about it. Binning is used for video as Full HD movies have a much smaller resolution than the sensors. And even if the D800 is supposed to have a better ISO performance than the 5D Mark III (at least for still images, according to DxoMark), for video it's acutally .

Test : vimeo.com/40113110

[Manfred M]

Francois - the first thing I bought after I got my D800 about a year ago was a 4TB hard drive to store the data. While it was overkill, but I got it for a great price. Storage is cheap.

I would need a lot less storage if I were more aggressive in culling questionable images, but I tend to not do so unless there is a really major problem with it.

[Francois]

You're right.

Even more because I store lot of my images in jpg, and only keep in raw the one I like the most, or the one of important moments like a wedding.

But the first concern wasn't storage cost, indeed cheap, but trying to get the best quality possible when I don't need huge pictures.

[revi]

Remco; you are absolutely correct in what you say; but you can still get a result that is "good enough" working from a binned image. If storage space is a priority, it will give adequate results; after all, I have gotten some stunning images that I edited out of 4MP and 5MP point & shoot cameras that output jpegs only.

Oh yes, I agree. You don't need 36 MP (or even 9MP) in a lot of cases. 9MP is already an A4 print @300 PPI, so should be more than sufficient for any size print, and way too much for current screen sizes.

But I thought François' question was, what was the best way to get a 9MP image from the original 36MP raw file. He suggested binning each bayer array to get one pixel. And I'm not sure that's better than go the 'standard' way of full-size demosaicing and then resize. Otoh, I'm not sure the binning method is really wrong, either (I suspect it's what's done to get the jpegs embedded in raw files, with a lot of other processing, as it is fast). With current software it's just a lot of hassle for no gain (and potential loss of quality).

[ajohnw]

Binning is usually a function of the sensor and done with hardware. Not at all sure it makes sense on colour sensors as it might involve treating 4 pixels as 1 for instance. Size reduction software can have all sorts of things going on in it.

John
-