Cameras with RGB/Bayer colour sensor?

[Wayne Morellini]

Hi,

I looking for video camera models that use a RGB/Bayer colour filter on their sensors, instead of the normal complementary colour schemes? I have heard that there are a few cameras with them.

The reason why is they give truer colour rendition.

[Philip Williams]

Originally Posted by Wayne Morellini

Hi,

I looking for video camera models that use a RGB/Bayer colour filter on their sensors, instead of the normal complementary colour schemes? I have heard that there are a few cameras with them.

The reason why is they give truer colour rendition.

All the current Canon Opturas use primary RGB color filters. I think some of the Sony 1 chip camers do as well.

Philip Williams

[Wayne Morellini]

Thanks for that Philip.

Does anybody else know of any other models available to compare to them?

[A. J. deLange]

One chip cameras (Foveon based excepted) must, of necessity, have a Bayer filter or something very like it (some of the still cameras have begun to use colors other than or in addition to the traditional R, G and B). I question the premise that they produce better color though. If they did the most expensive cameras would use them. In fact it's just the other way round.

[Tommy James]

The state of the art for one chip camcorders is the hybrid primary complimentary color filteration system. My JVC JY-HD10 camcorder has this technology.

[Wayne Morellini]

Thanks A.J and Tommy.

Most cheap cameras use a complementary colour system. This is where they use complementary colour (each a mixture of two primaries). Unlike primary filters, this only filters out one primary colour per sensor pad (depending on scheme) allowing more light.


We see in primaries, and the complementary system has to convert it back to primaries to record on tape. But the complementary scheme does not actually tell you how much of each primary is at the complementary sensor pad. So they have to make an educated guess as to the real value, this reduces the accuracy and purity of colour (unless they raise saturation, reducing colour accuracy again). After this it has to work on the missing colour in the location. So you are guessing for all colours with a 4:2:0 scheme this does not matter as much as much colour is destroyed. What is important is the intensity of each pixel, but the intensity can also be effected by this scheme. This leaves to single chip camera of this sort being deemed unsuitable for TV work (but they are known to accept anything to make a buck) because it produces strange effects and clouds etc.


With Bayer (RGB+another Green often used in good digital still cameras) things are different, one colour per location is guaranteed, so accuracy is starts off better, intensity and the other missing primary colours per location are worked out from the accurate colours of the surrounding pixels. More expensive video cameras/professional Eng camera often use three chip that shoots each primary to a different sensor for more accuracy, but digital cinema cameras, used to make movies, often use the bayer scheme because it allows them more freedom in quality and types of changeable lens, over three chip. The Foveon X3 is an attempt to put 3chip in one chip.

There are a number of primary schemes RGBG, BRGB, RGB etc. But remember, on a cheap camera there are many ways to "dumb" down the image for the consumer market, even if it has a primary filter. In our own digital cinema camera attempts (and I mean attempts, the cheapest is around $17K at the moment) over at alternative imaging, we have shown it is possible to get a much better bayer single chip image compared to most video cameras.


The hybrid scheme, used in the HD1/10 produces nicer colour than the complementary schemes I have seen, but this is deceptive in itself. Though, I must say the pattern is is ingenuous. I have heard the camera has minimum gain due to problems of noise it produces in the reds which I have heard is marginalised. I have also seen much colour noise in footage in reasonably lit low light environments and heard of the colour going murky past that (but that also happened with auto colour control on the GRDV3000 too). But looking at the filter pattern itself, something else became apparent, it uses a white tile (all the colours). If this tile is clear then the situation point of that pad will be much lower than the surrounding pixels. If it is opaque then the saturation point can be leveled with the other colours. The green tile has the opposite problem unless all the other tiles light transmission is matched. The cyan tile is curious, if a true mixture of green and blue then OK. I'm conjecturing here, I haven't access to the accurate data, Steve Mullen might know. So why not just even everything out, I think because it would hinder low light performance even more. Having the white tile clear and a t maximal transmission, and the complementaries at maximum light transmission would help get some image in low light, even if it is uneven. But bright spots in the image will go first, but the white tile can be guesstimated in, who this would effect the image I wouldn't know. Now the cyan, is it Blue green, or more green. For video, and in human vision in some ways, green has the most accuracy/detail, and closer to yellow in another way (I don't remember the details). So more green is better.

The three other negative things I have heard about the image of the camera that uses this hybrid scheme are: Not enough latitude (the range of brightness between the brightest and dimmest details it can record) in bright light, reds having noise/less accuracy (I think), and flames of matches being lit in dark places going green (I assume will turn up elsewhere, plus I have observed very bright colours are much brighter compared to other colours. In s bright situation, using no gain should give you more latitude. Maybe the tiles are using different light transmission. This should effect the reds as well, due to it being in the yellow and white tiles and maybe the normal consumer video accuracy problems of red. With blue it doesn't matter so much, because our eyes discern the least amount of levels of blue compared to any primary (green the most). Because of the way the scheme works and the latitude problem, I expect this is the reason that the bright colours saturate out (looks trendy though). With the flame, maybe this is caused by too much green sampling causing the camera to become confused temporarily about all the light it is getting on all tiles (I can guess it settles down eventually). This is only conjecture though, people with this sort of camera should be able to work it out themselves.

I think the JVC hybrids ares fine in well lit situations.

I should have stated, with enough pixels (usually at least 680K to 1.3Mp) colour/accuracy seems to increase with a complementary scheme (at the expense of light gathering power from the extra pixels).

[Wayne Morellini]

Back to cameras.

I've looked through most of the cameras suggested, but I think nearly all are around 680K pixel interlace.

I should have stated I am looking to do an experiment and maybe even make it into a product. But I now work out I need a camera of at least 1.3 Mpixel with an progressive scan mode. I might have to wait to see the new JVC hard drive HD camera.

[Philip Williams]

Originally Posted by Wayne Morellini

Back to cameras.

I've looked through most of the cameras suggested, but I think nearly all are around 680K pixel interlace.

I should have stated I am looking to do an experiment and maybe even make it into a product. But I now work out I need a camera of at least 1.3 Mpixel with an progressive scan mode. I might have to wait to see the new JVC hard drive HD camera.

Probably not useful, but the Optura can be set to "photo" mode and it then appears to stream a 4x3 progressive video through firewire. CCD is around 2MP.

[Wayne Morellini]

Interesting, we were having a discussion somewhere else about some cameras put a higher quality picture out the video port during still camera mode, and maybe the firewire. Thanks for the tip, I'll have to look it up again. But if there are any others I would like to hear about them too, to make the best cheap choice.

Do you have confirmation of the quality of image from firewire, format, and compression ratio? In truth I am searching for a super cheap way to get quality HD to people from conventional cameras, but I am having trouble finding a suitable candidate camera, that's why I am asking whats, most have at least one serious problem.

Thanks for your help.

[Philip Williams]

Originally Posted by Wayne Morellini

Interesting, we were having a discussion somewhere else about some cameras put a higher quality picture out the video port during still camera mode, and maybe the firewire. Thanks for the tip, I'll have to look it up again. But if there are any others I would like to hear about them too, to make the best cheap choice.

Do you have confirmation of the quality of image from firewire, format, and compression ratio? In truth I am searching for a super cheap way to get quality HD to people from conventional cameras, but I am having trouble finding a suitable candidate camera, that's why I am asking whats, most have at least one serious problem.

Thanks for your help.

You know, I never did see if a progressive video stream was coming off the SVIDEO port when I tested with my Optura 30. The firewire output was, of course, just 720x480 DV. However it did have noticably more detail than the interlaced footage recorded to tape (and that was with a locked shot). I seem to recall that some off-angle line detail had a stair step effect that may have been created when the stream was downsampled. I actually cropped and resized some of the frame grabs to 1280x720 and they held up surprisingly well. I bet if someone could bypass the DV compression on an Optura (like the real-stream with the DVX) you'd get some impressive video. I never got very far with my own testing though. I was just looking to get quality progressive scan, but the fact that I'd have to lug a laptop around to record the live stream - combined with the emergence of lower cost HD around the corner - kinda burried that idea.

[Wayne Morellini]

Originally Posted by Philip Williams

You know, I never did see if a progressive video stream was coming off the SVIDEO port when I tested with my Optura 30. The firewire output was, of course, just 720x480 DV. However it did have noticably more detail than the interlaced footage recorded to tape (and that was with a locked shot).

This is what I meant about format. If it is straight 25mb minidv then you wont get any advantage, except maybe the tonal/light/range values of the image maybe better. But if the stream had more bits per colour, 4:2:2, lower compression, different compression format, progressive that would be worth looking at.

I seem to recall that some off-angle line detail had a stair step effect that may have been created when the stream was downsampled.

That's the problem, it's hard to tell what is improved. With the 35mm adaptor the blurring of the pixels looks better to some people, because it makes it look smoother by hiding the pixels. It might be less interpolation, or something else designed to effect preview quality.