Fractions of a pixel?

[Rufus]

I have always wondered how there can be fractions of a pixel. For example in feathering and sharpening some variables can be set to a decimal place of a pixel. This is counter-intuitive to me as I thought a pixel is the smallest element of a digital picture file or a screen display.

I am sure that knowing the answer is not going to improve my photography, but it I am just curious about this nagging question!

[Manfred M]

A pixel is a discrete unit and is dependent on the camera's sensor. Nothing suggests that you can't do / resolve things at a sub-pixel level.

When I do my sharpening, I will often apply it at a sub-pixel level. Not only does it do its job, but a small radius reduces the incidence of halos.

[dje]

David

With most processing filters, some sort of mathematical distribution function is used to calculate multipliers to be applied to pixels around the centre pixel. For example, with the Unsharp Mask filter, a Normal (Gaussian) distribution is used as the blur function. The "pixel radius" setting relates to the standard deviation of the function which is an indication of how broad the function is. See the different colored lines in the following diagram.



If you imagine that the integer values on the X axis are the positions of the centres of the pixels, then the values of the function at those points are used to calculate the multipliers for those pixels. As the distribution is a mathematical function, the standard deviation is not limited to integer values.

Make sense? This is the just the sort of thing involved.

Dave

[pnodrog]

I think it's a bit like currency. I see a lot of items priced at $1.99 when the smallest amount of change they can give is 5 cents.... (that's my two cents worth anyway)....

P.S. I see Dave has entirely confused the issue by giving a logical explanation...

[DanK]

Dave,

Thanks. that does make sense. I think what causes some confusion is that the radius in sharpening algorithms is often described as discrete, as if a pixel is or isn't changed by sharpening. Were that the case fractional pixels would make no sense.

Dan

[xpatUSA]

I have always wondered how there can be fractions of a pixel. For example in feathering and sharpening some variables can be set to a decimal place of a pixel. This is counter-intuitive to me as I thought a pixel is the smallest element of a digital picture file or a screen display.

I am sure that knowing the answer is not going to improve my photography, but it I am just curious about this nagging question!

It happens with 'bits' too! For example, analog-to-digital converters have a measure of performance "effective number of bits" often expressed to two decimal places.

And then there's "bits per pixel" an inverse measure of JPEG compression which can be more than 8, oddly enough.

Not forgetting quantum efficiency - how many indivisible photons get captured as electrons by the sensor on average, which is a fractional number, albeit usually expressed as an integer percentage ...