Best camcorder for stills

[PeterReynolds]

The Sony PC100 is currently my favourite of several camcorders I have owned and 50% of my usage is for hi-res stills on a Memory stick.

Would the new JVC GR-DV2000 improve on the quality of stills I get from the Sony PC100?

Peter Wells does not mention this part of the spec in his generally favourable review in April's CV (page 58) but one Jim Tyler in sister magazine CU (March, p 20) describes a pixel-enhancing process which delivers 1.92 million pixels for an image measuring 1,600 x 1,200, all from a source CCD of 800,000.



------------------
Peter

[Alan Roberts at work]

You'll find that all digital cameras use pixel enhancement techniques to produce images with apparently greater resolution than the sensor should permit. There's a lot of good science behind it, and none of the manufacturers has any significant advantage over the others.

E.g., my Olympus C3030 has 3.34M pixels in the ccd (it says so on the box) but it makes images at 2048x1536, and can make a TIFF file uncompressed, so they're not afraid of people like me having a darned good look at how they do it. That image is over 3M pixels, but is RGB in the file; to get that resolution without trickery would need either 3x3.3M ccds (for RGB) or a single 9.9M ccd. So, clearly, there is trickery going on. I was at a conference a couple of years ago when some detail of these processes got out, and believe me, you'd be amazed at the amount of processing done in even the simplest of digital cameras.

Video cameras do less trickery because they have to do it at video rates, and that means faster processing. Nevertheless, there are still clever processes going on in video cameras to produce apparently higher resolution than the pixel dimensions of the imager would allow.

There are some very clever people at work in this industry.

------------------
alan@mugswellvillage.freeserve.co.uk. Delete village for a spam-free diet.

[PeterReynolds]

Is there, then, any objective way to compare the stills capability of camcorders?

The Sony PC100 stills which I like come as .JPG files of less that 500kB.


------------------
Peter

[Pierluigi]

Alan,

Any chance of writting a brief description (if posible) of what single CCD devices do to get the resolution that they claim and how that differs from 3CCD models?

Many Thanks

Lui

[Alan Roberts at work]

Oh dear, I wondered if that would happen. I didn't take any notes (I was setting up lots of kit ready to deliver the next paper and only heard it in the background) so I can't be accurate. But I know it involved what tv calls "aperture correction" and Photo Shop calls "unsharp mask" (they do the same thing) and a lot of edge detection and enhancement. But the most important aspect is to get the mosaic pattern right, the layout of the RGB sensitive pixels. A lot of work has gone on in finding optimal layouts, based on the eye's poorer spatial response to blue and red than green. This comes from the eye having cone cells in the ratio of about 40:20:1 (G:R:B) and the fact that it then does some spatial filtering in the retina and codes the data into a form of YUV with lower bandeidth in the colouring signals than in luminance (note that here I talk of "luminance" and not "luma", because the term is correct here, tv does not deal with luminance but a wrongly coded signal that masquerades as luminance which I and some others call "luma").

All this processing is done on the gamma-corrected signals in the camera, and that gives rise to some profits and losses in the area of silicon because some things are easier done in non-linear but aren't as effective.

And so it goes on, there's lots of it but I can't remember any more precisely. I'll dig about and see what I can find. The best text book I can recommend for anyone interested is by R.W.G.Hunt, called "The Reproduction of Colour" now in about 6th edition. It's big and expensiove but very good. He was Assistant Director of Research for Kodak in the UK for over 30 years, and Prof of Colour Science at City University for many years, senior lecturer in the colour courses I attended about 15 years ago. He still lectures at CII in Derby University (Colour and Imaging Institute). Anyone seriously wanting the learn this game should go on one of their courses, I think one has just completed (the Easter one). The alternatives are to get onto metrology courses at NPL or City University, I've done most of them and learned hugely from some of the world's best experts. Sitting for 90 minutes listening to and watching Bob Hunt is both a delight and an education, he breezes in with a carousel of slides and no notes and talks for exactly 90 minutes, finishing on the button. It's sheer magic, the best educational fun you can get someone else to pay for (if you like that sort of thing).

Sorry, got to go now, a meeting in a big building in London to sort out someone else's problems (that seems to be all I do these days).

------------------
alan@mugswellvillage.freeserve.co.uk. Delete village for a spam-free diet.