Using less pixels to decrease noise

[Daniel Smith]

Hi.

 

I don't know how more capacitors on a CCD creates more noise however, a lot of cameras, mainly stills cameras offer the option of shooting in a lower resolution.

 

Do you think it takes the photo at full resolution and then down-rezzes it or will it use less sensors on the CCD? And, will using less sensors on the CCD have any effect on noise, even amongst the millions more un-used?

 

cheers.

Dan.

[David Mullen ASC]

Hi.

 

I don't know how more capacitors on a CCD creates more noise however, a lot of cameras, mainly stills cameras offer the option of shooting in a lower resolution.

 

Do you think it takes the photo at full resolution and then down-rezzes it or will it use less sensors on the CCD? And, will using less sensors on the CCD have any effect on noise, even amongst the millions more un-used?

 

cheers.

Dan.

 

I'm not sure, but the cameras I've used in the past store lower resolution files as JPEGS, so they are downrezzed RAW conversions, not cropped sensor images.

 

I don't know if some DSLR's actually store RAW / NEF files in different resolution sizes without cropping the sensor area, since that would require RAW to RAW rescaling.

 

I would think using less pixels on the sensor would not affect noise, but once you enlarged the cropped image to the same size as the full sensor image, and thus saw less resolution, the noise artifacts may actually become more visible. At least, that's been my experience in HD movies with cropping and enlarging frames -- you're better off oversampling than undersampling.

[John Sprung]

Do you think it takes the photo at full resolution and then down-rezzes it or will it use less sensors on the CCD?

If the field of view for a given focal length stays the same, it's using the full sensor and downconverting. If it crops the sensor and only uses part of it, like the Red 3K and 2K options, it's like going to a smaller negative in film. The same lens appears longer.

 

Oversampling and downconversion is always better. Noise is just random numbers added to or subtracted from what the real value should be for each pixel. Downconversion reduces noise because the additions and subtractions tend to cancel each other out somewhat. But there's no free lunch, it also reduces resolution.

 

 

 

-- J.S.

[Paul Bruening]

I was under the impression that post-capture, pixel averaging was the most common system for down-ressing. The bum pixel gets stirred in and reduced with three other pixels. Decent cameras have selectable noise reduction on board which uses some of the same neighboring pixel averaging in both high and low-res.

[Scott Fritzshall]

Having more photosites on a sensor of a given size tends to increase the noise because it means that each individual photosite is smaller, and therefor receives fewer photons, reducing its sensitivity. Digital stills cameras offer different sizes simply for storage reasons; when your mom is out taking pictures and doesn't really care about them having a zillion pixels, she might want to shrink them down in order to fit more of them onto a card. This only affects the processing done to the RAW image after it's already been taken; it doesn't change anything about the input image. Now ultimately, doing this will probably reduce noise, but that's just a side effect of downsampling, rather than anything to do with the sensor.

[Daniel Smith]

Ok tnx.

 

I was thinking about the HDX900 also, which offers both HD and SD, it would be nice to record some SDI on a 2/3" as the DSR-570's and 500's we have don't have an SDI out.