Bayer Begone

Discussion in 'Mirrorless Digital Cameras' started by rayfraser, May 28, 2005.

  1. Success on my fourth attempt. Bayer is gone! Small SMaL photo follows links to previous attempts and a full image plus RAW (1284x864 pure B&W with no interpolation or post processing). Note I realize I need to take it apart again and cleanup a few specs.
    Bayer Scraping - Initial Result
    Chemicals in Digital Darkroom - B&W Debayer Success
    Full image
    I would appreciate any suggestions on how to demonstrate higher resolution capability of noninterpolated-unencumbered pixels. Obviously a black and white clock is a poor choice.
  2. Looks like you're getting there, Ray. Keep after it.
  3. Are the letter and number shadows (easily seen on linked full image) due to my shaky hands? Or was it flash through glass cover? Letters and numbers don't look raised - so that along with blurry backwards digital reflection certainly means I was shaking. Clearly detailed comparison shots against a bayered imager will require a tripod.
  4. I see shadows only on certain numerals so they appear to be actual shadows, not motion blur due to shake.

    It would be better to photograph something like black figures against a flat white background without shooting through glass or plastic. The glass or transparent plastic is likely to distort the black characters, making it more difficult to evaluate results.
  5. Ray! Will you post a step-by-step HOWTO with pictures?

    What software did you use to process the RAW image into a JPEG?
  6. More to come, but I need some Zs. Image below is cropped from raw unprocessed pixels using IrfanView. Top was taken with a CVS typeid 2F, bottom was with a CVS debayered typeid 2F (note 2Fs have a newer larger imager than older 28s). I had to refocus factory set infinity on debayered - I think due to removing glass cover over imager (that could invalidate results). By zooming in you should be able to see lower valued pixels on top. I will make full raws without headers available if anyone wants to become a debayer player (first confirm capability by processing clock RAW). I will follow this post with another showing uncropped images processed using FlatFoto software.
  7. And here are originals resized to fit 511 limit.
  8. Now you're getting somewhere, Ray! I'd enjoy seeing some more samples, especially outdoors... I can't quite see the advantage so far :) It's very difficult to tell at the size, but the top (with bayer?) looks sharper than the bottom (without bayer?) in the close-crop picture of the road sign.
  9. There's no reason why the "debayered" image should be sharper than the "bayered" image. They both have the same number of pixels.

    If you wanted to demonstrate superiority, you should shoot with the regular sensor, then do a color -> B&W conversion using only the red channel. Compare that with an image shot with a "debayered" sensor using a red filter over the lens.

    In principle the "debayered" image should then be better since it would be using 3x as many pixels (assuming an RGGB bayer matrix).
  10. I have taken over 30 pictures with each camera and the one above seems to have best demonstrated a higher resolution - if you download the tiny crops and blow them up you can see dark low value in every other pixel in green area on sign for bayered image. The bottom has what may be perceived as a sharper cleaner look because it lacks the darker pixels under blue & red filters in green area. In these cameras as well as others using normal bayer color filters every other pixel has a green filter and reds and blues are used to fill the gaps on alternating rows - thus Bob's suggestion would result in 4 times as many pixels. I think that only in very special situations such as using a color filter will debayering make an obvious difference. If the polymer filter say for blue was ideal and only let blue light through and a blue sky only had blue light then I would have posted some of my clouded sky comparisons.<p>
    One of my long term goals is to experiment with UV shots and removing bayer seems to have let another stop of light through (these cmos electronic rolling shutter imagers have a max exposure of 1/30 second). I also want to test my feeling that extended dynamic algorithms may work better when all pixels respond only to intensity without alternating color filters adding zoning confusion.
  11. Sorry I almost was deconvinced about debayering after reading Bob's response. It wasn't until long after responding as I was drifting to sleep that I wondered about how comparison images would appear on a B&W print or even a old monochrome monitor. I think as you blow image up on a color monitor it becomes obvious that raw debayered image provides more detail (see image below - I think the darkest pixels are under red, but maybe those are blue filters).
    As to a "How to" I still am not sure what if any chemicals are necessary. After rubbing with acetone and phosphoric acid (nail polish remover and aluminum jelly), and not seeing any positive results I have grabbed xacto knife which was at my side having used it to cut glass cover off and used it to start peeling bayer. It goes too deep and on my latest attempt I just used it to start in upper left corner. After that I scraped the rest away using broken plastic toothpicks (fresh breaks scrape faster). Color changes to gold where thin bayer has been removed making it easy to rapidly progress around imager (be very careful to avoid gold bond wires).
  12. "One of my long term goals is to experiment with UV shots and removing bayer seems to have let another stop of light through (these cmos electronic rolling shutter imagers have a max exposure of 1/30 second)."

    Feel free to contact me if I can be of any help.
  13. Thanks, Vivek, I will contact you. First I need to obtain a tiny 9M or 12M board sized quartz lens (also considering pinhole possibility but I doubt if tiny pinhole blocked to all but UV would let enough light through in 1/30 second).<p>
    To be fully honest my above B&W cropped sign comparison used raw pixel values to emphasize differences. If nearby colors were interpolated and merged into each pixel the differences would only appear in areas of rapid single color change.
  14. You are welcome, Ray. A detailed email is on its way to you, later today.

    I do not know for sure but I think the UV sensitivity of a CCD or CMOS sensor is not lower than 350nm (below 400nm being UV). Even if a CMOS sensor has UV sensitivity below 300nm, it is of no use as practically no UV gets through anywhere in air. One needs high vacuum for far UV (below 300nm) to avoid loss due to scattering.

    Given this, if the lens on the $10 camera, even if it is glass (most likely it is acrylic plastic which is OK as well), it should have enough transmission for your sensor (with or without the Bayer).

    So, all you need is a bandpass filter that cuts off all visible light but is transparent to the UV. I would suggest Hoya U-360. This is the cheapest and most versatile of all the bandpass filters currently on the market. A 12mm circular filter should do for you, I hope. Filter is more expensive than the camera you are working with!

    For availability and pricing:

    I would start experiments on a very sunny dayfrom 2-3pm. Point at the shaded area to make test pictures. There will be enough UV in the shade compared to visible and infrared light.
  15. Nice work - as far I know, you are making history. There have been numerous debates over removing bayer filters in astronomy oriented groups, but this is the first time to see actual successful implementation. May I ask for closeup picture of the sensor - I am interested in specifying the manufacturer, it may be possible to use this approach to other sensors. Theory says, that using right activators to make the agent binding color dye on sensor photosensitive again for dye bleaching or binder dissolve should make possible the removal without mechanic stress to surface. There are numerous patent descriptions up in web, but the methods creating bayer filters seem to be significantly differing among chip manufacturers.
  16. This is a an old thread, but I still find it interesting.
    We have been doing some work at removing the Bayer filter array from a variety of Nikon and Canon cameras. Nothing for sale yet, but you can see some of our testing and examples of a stock camera and our modified camera.

Share This Page