How a sensor would work in the fuji sl/s series

Started Jan 9, 2013 | Questions thread
OP andy10 Regular Member • Posts: 233
Re: 'EXR-SN' versus conventional 'down-sampling'...

Mark H wrote:

andy10 wrote:

I currently own the HS20, which I use mainly in EXR dynamic range mode. Sometimes I use the EXR SN mode where the sensor combines adjacent pixels, so that each of the larger 8m pixels lets in twice as much light. The new SL1000 and other new S series cameras have 16m pixels available, and no EXR modes. If you set the image size to 8m pixels in these cameras would it combine adjacent pixels to increase the light per pixel, or would it simply cut out half the pixels to reduce noise?

In your later (non EXR) case, the camera doesn't really do either as you describe.

In a non EXR camera, a half resolution image (e.g. 8MP image from 16MP sensor) is simply produced by first creating the full 16MP image as normal, then 're-sampling' the image in the same way that any image processing software would do on your PC.

The only real advantage to having the camera do this, instead of later on your PC, is that the resulting image will take up less memory storage space, the sharpening/noise reduction etc might be better optimised for the smaller image size, and there might also be some slight saving from not having to JPEG compress the image twice or more as a result of resizing in PC post-processing.

The advantage of the EXR-SN process is slightly better still - because the down-sizing is achieved by combining adjacent sensor pixel's signals before they are A/D converted. By combining two pixels signals this way, the signal is doubled, so the 'ISO amplification' is halved and therefore the 'read noise' is halved - also, it makes all the cameras nearly twice as fast because it is only converting/processing 8MP, not 16MP, of digital data.

Halving the read noise is a real advantage, but isn't 'quite' as big an advantage as it first sounds, because conventional 're-sampling' down from e.g. 16MP down to 8MP, can effectively reduce the same noise by 30% anyway just due to the statistical averaging.

Finally, 'read noise' only really dominates in the shadow tones of an image - but it's clearly a significant issue as ISO increases, so even the fractional advantage provided by the EXR-SN mechanism is certainly very useful at higher ISO, particularly so in smaller sensors.

This reply is very helpful, thanks


Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark MMy threads
Color scheme? Blue / Yellow