|
|
Posted by Dave Martindale on 11/09/07 02:12
Jan Panteltje <pNaonStpealmtje@yahoo.com> writes:
>It is a bit more complicated then that.
>This camera, the HV20, is a single sensor chip model.
>To get color, the chip's sensor elements are covered by red, green, and blue
>filters:
> http://en.wikipedia.org/wiki/Bayer_filter
>Likely Bayer patterns (as Canon mentions 'simple RGB' in their specs).
>Just this fact, in itself, sort of reduces the luminane resolution by
>half the horizontal pixels, see:
> http://www.cambridgeincolour.com/tutorials/sensors.htm
No, that's not correct. A Bayer-sensor camera is able to achieve a
*luminance* resolution that is comparable to a monochrome camera with
the same number of pixels, not half. The reference you point to even
says that you can do better than half resolution in demosaicing.
In the digital still camera world, which I'm more familiar with, a 6
megapixel camera with a sensor that's about 3072x2048 pixels can resolve
nearly 2000 tvl per picture height - not the half-resolution 1000 that
you suggest.
The chroma-only resolution (ability to resolve features that differ in
hue and saturation only, not luminance) *is* degraded by using a Bayer
sensor, and it's generally downsampled by 2 before encoding anyway.
But picture content that has only chroma differences is rare unless
you're shooting a test pattern designed to show it. And the human
eye's chroma-only resolution is *one tenth* as good as its luminance
resolution, so recording images where the chroma is merely half
resolution is more than good enough.
>This is a consumer camera.
>Canon mentions optical stabilisation, but in many consumer cameras there is also
>digital stabilisation, the way that is done is the picture is projected on
>a smaller area then the full sensor, so it can move about horizontally
>and vertically,
>allowing for the electronics to move processing along with it, resulting
>in a stable
>picture.
>Such a scheme further reduces available resolution.
>I am not sure if the HV20 also has some digital stabilisation.
It's unlikely, since optical stabilization is superior. Digital
stabilization merely removes longer-term movement between frames, while
leaving the effect of blur *within* one frame period. Optical
stabilization also deblurs individual frames.
> http://images.camcorderinfo.com/images/upload/Image/Canon/Canon%20HV20/Video_Performance/3000lux/Canon_HV20_3000lux_24P_vivid.jpg
>quote:
> In 1080i, the HV20 showed a vertical resolution of 575 lw/ph and a
> horizontal resolution of 625 lw/ph. In 24P, the camcorder actually
> improved the vertical resolution, producing less break-up and
> artifacting in areas of high-density information, boosting the vertical
> resolution up to 600 lw/ph. The horizontal resolution remained unchanged
> at 625 lw/ph.
>end quote
I don't see how some of those numbers come from that image. Look in
the upper left corner - the vertical resolution test pattern resolves 5
bars all the way to the end of the pattern at 800 tvl/ph. The camera may
be capable of more; the test pattern just isn't capable of testing it.
(Calling horizontal resolution 625 tvl/ph seems fair).
And that's measured in the corners and edges; the resolution might well
be better in the centre if the test target had a suitable test target
there.
>But it gets worse, playing the 1440x on a true 1920x1080 LCD reduces
>detail further
>due to re-sizing in that LCD (or call it aliasing) as 1920 / 1440 =
>1.333.. and every so
>many pixels some artefact will appear as unsharpness (I should take a
>picture of this effect
>some day, but it is clearly visible on small text, also on moving fine
>patterns).
There's no further loss of resolution necessary. If horizontal
resolution is 625 tvl/ph, that's about 1100 tvl per picture width.
That can be stored in 1440 samples/line without significant further
loss, and 1440 samples can be upsampled to 1980 samples without
any visible loss. Now, some TVs have lousy interpolation algorithms,
but that's not the camera's fault.
Dave
Navigation:
[Reply to this message]
|