Thursday, June 14, 2012

EDN Discusses Image Sensors for Embedded Vision Systems

EDN published large 2-part article on image sensors for embedded vision systems. 6-page Part 1 is titled "Image sensors evolve to meet emerging embedded vision needs" and is written by Embedded Vision Alliance editor-in-chief Brian Dipert and his colleagues Eric Gregori and Shehrzad Qureshi. In a very popular form it describes automotive, mobile and regular cameras, 3D imaging and light field cameras.

The second part is devoted to HDR imaging and is written by Michael Tusch, CEO of Apical. Talking about various HDR implementations, the article describes Altasens' 1080p/60 A3372 sensor using a "checkerboard" pixel structure, wherein alternating Bayer pattern (RGGB) quad-pixel clusters are set to long- and short-exposure configurations:


"Long exposure delivers improved signal-to-noise but results in the saturation of pixels corresponding to bright details; the short exposure pixels conversely capture the bright details properly. Dynamic range reaches ~100 dB. The cost of HDR in this case is the heavy processing required to convert the checkerboard pattern to a normal linear Bayer pattern. This reconstruction requires complex interpolation because, for example, in highlight regions of an HDR image, half of the pixels are missing (clipped). An algorithm must estimate these missing values.

While such interpolation can occur with remarkable effectiveness, some impact on effective resolution inevitably remains. However, this tradeoff is rather well controlled, since the sensor only needs to employ the dual-exposure mode when the scene demands it; the A3372 reverts to non-HDR mode when it's possible to capture the scene via the standard 12-bit single-exposure model.
"

Update: Another version of this article is published by EDN-Europe on Oct. 1, 2012.

10 comments:

  1. Dual exposure is problematic for video, movement artifacts can appear.
    I'd rather see more color filters in the array, something like RGBWK or RGBrgb.

    ReplyDelete
    Replies
    1. I think you misread the piece about. AltaSens' approach is to simultaneously capture long and short exposures using a vertically and horizontally interleaved checkerboard pattern.

      While each field will experience different amounts of motion blur, at 60 fps that shouldn't be much of an issue.

      Delete
    2. The piece ABOVE, not ABOUT.

      Delete
  2. Nice twist on an old idea. I guess noise and resolution are a bit unusual after ISP in bright and dark areas, but ok under lighting that falls in the range for both. But, I don't think it has any advantage except for memory requirements and IP over other older approaches.

    ReplyDelete
  3. I think that SONY has a patent on this idea. They called SVE = Spatial Variable Exposure. Please check it. The exposure difference can be created either by filter transmission coefficient or by using different exposure time.

    ReplyDelete
  4. I tried and no success. So, I doubt your information.

    ReplyDelete
  5. Check this one out - US #7719573.

    ReplyDelete
    Replies
    1. This seems to be a CCD patent and does not seem to anticipate CMOS image sensors in the specification. Strange, because it seems to date from at least 2003.

      It does describe an "SVE" concept but sure seems like a nuisance to implement with a CCD!

      Delete

All comments are moderated to avoid spam and personal attacks.