Leftovers from last lecture

  - number of bits useful in digitizing output from linear arrays
      (such as those found in flatbed scanners)
     - a random example: Kodak KLI-14403 (3 x 14404; 5um)
        - depth 250K e-
        - dyn range 82 dB (20 e- read noise) -- 13 or 14 bits useful
        - well depth larger than for comparable 2D CCD.  Why?
        (note that this is a fancy new sensor)

  - Nayar & Mitsunaga's SVE interpolation scheme (copied from last lecture)
     - easier than CFA b/c recovering one function, not 3
        (they don't address color -- they have RGB per pixel)
     - their approach
        - define an interpolation from inter-pixel grid to pixel grid
        - for each off-grid pixel:
           - look at 4x4 neighborhood of on-grid pixels
              [16x1 vector M_r]
           - look at 7x7 neighborhood of off-grid pixels that influences M_r
              [49x1 vector M_o]
           - consider the linear transformation defined by interpolation
              [16x49 vector F : M_r = F M_o]
           - solve for the 7x7 neighborhood of off-grid pixels that,
               when interpolated, will best for the valid on-grid pixels
              [First toss invalid rows from that equation, then solve
               M~ = F~ M_o using pseudoinverse]
           - take the center pixel from that neighborhood
              [So we really just need one 1x16 row of F~]
        - and then interpolate the whole off-pixel grid to get
            the final image
     - they don't discuss implementation
        - boils down to a 4x4 filter that depends on the validity pattern
        - one could imagine a lookup table -- 16 bits but 8fold symmetry...
        - maybe a table that only includes more common patterns...

  - Limits on HDR imaging
     - can only measure the dynamic range present on the sensor plane
        - lens flare
           - reduced by AR coatings, but still 1% bouncing around from each
               glass-air interface
        - camera flare
           - sensor is rather reflective; light scatters and eventually 
               reflects back onto the sensor
        - reduce both by shading the lens (only allow light in that will
            contribute to the image)
           - draw the stay-out region
        - reduce total energy by occluding unnecessary bright sources
     - keeping pixels separate
        - diffusion in substrate limits MTF
        - blooming is potentially a big problem
           - blooming defined; why it breaks HDR
           - HDR imaging through multiple exposures involves dumping a lot
               of photons on some pixels in the bright images
           - some chips have anti-blooming drains to collect overspill
               (costs fill factor)
           - the bad case is a bright pixel that has spilled into an adjacent
               one but not yet filled it up
           - could think of detecting it
               (if highest unsaturated value is not consistent)
 

Digital camera components

  - basic components are similar to film camera
     - aperture, shutter, lens, strobe, all with actuators
  - but some are different.

  - display
     - uses
        - viewfinder
        - UI
        - image review
     - advantages
        - accurate framing
        - focus preview
     - disadvantages
        - power hungry
        - low resolution (often judge focus by presence of aliasing)
        - preview useless in the dark

  - viewfinder
     - standard separate finder
        + cheap
        + sharp, real time
        - inaccurate framing for closeups (also impacts focus point)
        - sloppy coverage
        - no focus preview
     - SLR viewfinder
        + accurate framing (often accurate coverage)
        + focus preview
        - expensive
        - moving parts
        - prevents use of LCD viewfinder
     - SL/beamsplitter
        + simiar to SLR
        + no moving parts
        - dimmer
        - reduces light available to image sensor
     - microdisplay
        + framing, focus preview
        - low light
        - low resolution
     - eye relief defined

  - battery
     - need lots more juice
     - generally rechargable
     - borrow technology from camcorders (also very thirsty)

  - removable storage
     - microdrive (new but old)
     - flash memory
        - compactflash (driven by standard ATA disk interface)
           - pin-compatible with PCMCIA cards (=> cheap & trivial adapter)
        - other flash-like technology (smartmedia; memory stick)

  - I/O interfaces
     - serial (USB, 1392, IrDA)
     - video output
         (does anyone really use this?)
         (maybe popular among computer haters)
       

Digital camera processing architecture

  - processing system reads data from CCD, eventually produces JPEG
  - these days often a single-chip implementation (available from TI, Philips,
     Sharp, possibly others)

  - show block diagrams

  - autofocus
     - using conventional separate sensor
        - two ways: range finding or focus detection in image
           - (Marc discussed earlier)
        - can be faster
        - use additional information (e.g. uniformity over cone)
        - adds parts and complexity
        - generally only implemented in SLRs
     - using main image sensor
        - move focus and watch for contrast peak
        - slow to read out high-res device (increases shutter lag)
        - saves parts and complexity
        - high power to fire up processing chain
        - very flexible wrt. focus areas
        - algorithms
           - basically a servomechanism that uses the change in sharpness
              - sharpness as high-frequency content (bandpass then RMS)
              - sharpness as max of first derivative (maybe per row)
           - curve of sharpness vs. defocus
              - hope for one peak
              - multiple peaks from multiple objects
              - multiple peaks from contrast reversal
           - optimizations
              - macro mode vs. normal mode
              - last focus priority vs close/far priority

  - autoexposure
     - using separate reflected light sensor
        - technology from P&S film cameras
        - not that accurate, but cheap and low power
     - using matrix sensor at image plane 
        - technology from SLR film cameras
        - flexible and accurate
        - expensive: extra parts
     - using main image sensor
        - direct connection to ultimate image
        - slow (increases shutter lag)
        - high power consumption
        - lots of flexibility
        - illustrate with Philips alg diag