Forums

Chroma Decimation in Digital Video Signals

Started by Randy Yates October 14, 2005
glen herrmannsfeldt wrote:
> Chris Bore wrote: > >>> Often, the chroma data in the original content has already >>> been band limited to below 1/4th the sampling frequency. For >>> instance, in many digital cameras, a color mosaic in front >>> of the sensor pixels only allows sampling color at half of >>> the spacial sampling rate. > > >> I think a color mosaic only samples, it does not band limit. > > > In all video camera systems, as far as I know, it is the > optical system that does band limiting.
Heavens, no! I guess that by "optical system" you mean the lenses but not the sensors. The resolution of a decent lens is finer than the pixel spacing of most digital sensors that they're coupled to. I have used a photographic objective* capable of producing 1000 line pairs per mm on high contrast, high resolution film to probe the resolution limits of more ordinary film. Jerry _______________________________ * 50 mm f/4 by Wray of England. -- Engineering is the art of making what you want from things you can get. �����������������������������������������������������������������������
Jerry Avins wrote:

>> In all video camera systems, as far as I know, it is the >> optical system that does band limiting. > > Heavens, no! I guess that by "optical system" you mean the lenses but > not the sensors. The resolution of a decent lens is finer than the pixel > spacing of most digital sensors that they're coupled to. I have used a
Yes, that's why the lens of a digital camera should be designed to low pass (i.e. blur) the picture, in order to assure the Nyquist conditions for the sensor. Actually cheap cameras keep same lens with different sensors, so a 5Mpix camera could have same resolution as a 3Mpix camera. bye, -- piergiorgio
OK, this is interesting because you raise a topic on which I have found
it hard to get information.

So the camera lens can act as an antialias filter.
But only if the lens blur is similar to the pixel sample spacing?

So in fact using a better lens might introduce aliasing effects?

Are there any guidelines on choosing the lens to match the sensor?

Thanks,

Chris

Chris Bore wrote:
> OK, this is interesting because you raise a topic on which I have found > it hard to get information. > > So the camera lens can act as an antialias filter. > But only if the lens blur is similar to the pixel sample spacing?
Or more, as I mention, sometimes same lens are used for different Mpix setup, so you can get different sensors, but always the same "equivalent" resolution. Just to be clear: this is bad :-)
> So in fact using a better lens might introduce aliasing effects?
Yep. This is something that might sound strange, but that's it: better lens can mean worse picture.
> Are there any guidelines on choosing the lens to match the sensor?
You mean as customer of digital cameras or as manufacturer? AFAIK Canon has specific lens for their high-end digital cameras, even if I can immagine that a slightly unfocused setup will do the trick. bye, -- piergiorgio
I mean as a manufacturer.

I did in the mean time find information on optical antialias filters or
'optical low pass filters' (OLPF), which seem to be anything from a bit
of blurry plastic to a sandwhich of optically perfect crystal around a
phase plate.

Chris

Chris Bore wrote:
> I mean as a manufacturer.
Sorry, cannot help.
> I did in the mean time find information on optical antialias filters or > 'optical low pass filters' (OLPF), which seem to be anything from a bit > of blurry plastic to a sandwhich of optically perfect crystal around a > phase plate.
Uh, that's interesting. bye, -- piergiorgio
Jerry Avins wrote:

> glen herrmannsfeldt wrote:
(snip)
>> In all video camera systems, as far as I know, it is the >> optical system that does band limiting.
> Heavens, no! I guess that by "optical system" you mean the lenses but > not the sensors. The resolution of a decent lens is finer than the pixel > spacing of most digital sensors that they're coupled to. I have used a > photographic objective* capable of producing 1000 line pairs per mm on > high contrast, high resolution film to probe the resolution limits of > more ordinary film.
By the time the signal is electronic it has already been sampled. Thinking about it more, an interlaced scanning beam detector could average over about two lines worth, effectively filtering it. For CCD sensor that type of averaging is harder to see. The resolution you mention assumes the lens is properly focused at a large enough aperture to avoid diffraction effects. For real systems that may be relatively rare. -- glen
Chris Bore wrote:
> I mean as a manufacturer. > > I did in the mean time find information on optical antialias filters or > 'optical low pass filters' (OLPF), which seem to be anything from a bit > of blurry plastic to a sandwhich of optically perfect crystal around a > phase plate.
On the subject of lens design, lateral chromatic aberration needn't be a problem with digital color images. How odd that better color correction is needed for black-and-white than for color! Jerry -- Engineering is the art of making what you want from things you can get. �����������������������������������������������������������������������
glen herrmannsfeldt wrote:
> Jerry Avins wrote: > >> glen herrmannsfeldt wrote: > > > (snip) > >>> In all video camera systems, as far as I know, it is the >>> optical system that does band limiting. > > >> Heavens, no! I guess that by "optical system" you mean the lenses but >> not the sensors. The resolution of a decent lens is finer than the >> pixel spacing of most digital sensors that they're coupled to. I have >> used a photographic objective* capable of producing 1000 line pairs >> per mm on high contrast, high resolution film to probe the resolution >> limits of more ordinary film. > > > By the time the signal is electronic it has already been sampled. > > Thinking about it more, an interlaced scanning beam detector > could average over about two lines worth, effectively filtering it. > > For CCD sensor that type of averaging is harder to see. > > The resolution you mention assumes the lens is properly focused > at a large enough aperture to avoid diffraction effects. For real > systems that may be relatively rare.
That Wray lens is also relatively rare. It's glass diagram looks very much like a Leitz 100x dry objective, which has one eighth its focal length. The price if the Wray is just about 512 times the price of the Leitz. It is as if one buys optical glass by weight. Actually, it is 64 times the surface area that needs to be finished 8 times as well. Jerry -- Engineering is the art of making what you want from things you can get. �����������������������������������������������������������������������
It is suggested that natural scenes tend in fact to show less
high-frequency variation in color than they do in luminance. 'Natural'
being hard to define but still being a very small subset of all
possible images that one could imagine.

Similarly, 'natural' images tend to show more horizontal and vertical
lines than they do slopes, although this is not a very robust
suggestion.

And natural scenes tend to show more neutral colors than they do
extremes.

The logic for the human eye's having lower color than luminance
resolution is then turned back-to-front: the eye evolved that way
because nature is in fact that way.

Many images can be thought of, found in nature, and produced, that
contradict these suggestions - but often, such images turn out to look
'striking' - that is, they stand out just because they do look less
natural.

Funnily enough, man-made scenes also tend to show these characteristics
- for example urban rather than natural landscapes, although here there
are more straight lines and abrupt color transitions.

Chris
========================
Chris Bore
www.bores.com


Jerry Avins wrote:
> I was under the impression that much of the low-frequency nature of the > color signal arose in the eye, rather than the image. I can imagine that > that doesn't matter.