computational
play

Computational High Dynamic Range Photography HDR Frank Dellaert - PDF document

Computational High Dynamic Range Photography HDR Frank Dellaert School of Interactive Computing Georgia Institute of Technology Many Figures from Ron Brinkmanns Book Many figures from Debevecs paper Recovering High Dynamic Range


  1. Computational High Dynamic Range Photography HDR Frank Dellaert School of Interactive Computing Georgia Institute of Technology Many Figures from Ron Brinkmann’s Book Many figures from Debevec’s paper Recovering High Dynamic Range Intro Radiance Maps from Photographs Paul E. Debevec Jitendra Malik • HDR useful in many domains • Image = “brightness” • Rarely true radiance!!! • unknown, nonlinear, mapping On Black and White Sensor = Non-linear! 0-1 convention 255 = 1.0, not 255/256 ! digital brightness reduction does not do a good job

  2. Source of Nonlinearity Bracketing • Photographic Process: • Film Curve (base level, saturation) • Development/scanning/ADC • Digital Cameras: • Saturation, Bleeding • Re-mapping (12bit->8bit) Bracketing Bracketing Q: How do you expose detail in shadows ? Technique = Bracketing, E-split A: increase exposure time, increase aperture Lesson: white <> white !!! black <> black !!! Image Acquisition Bracketing Image Formation Irradiance Image E Z 1 =f(X1)=f(E ∆ t 1 ) (photons/sec) • Irradiance E • Exposure Time ∆ t Z 2 =f(X2)=f(E ∆ t 2 ) • Exposure X = E ×∆ t Exposure X (photons) = • Pixel Values Z = f(X) = f(E ×∆ t) exposure time ∆ t × E Pixel values Z = f(X)

  3. HDR Acquisition Estimating f E=f -1 (Z 1 )/ ∆ t 1 Average pM measured! Known! E=f -1 (Z 2 )/ ∆ t 2 1M 256 unknowns? unknowns? Caveats: • Known camera model f Taking the log makes time additive • Inverse of f does not exist for extremes • Treat superwhite/superblack differently My Iterative Procedure Estimating f, Debevec... • f -1: Z -> X is lookup table of size 256 • Guess f -1 (linear in certain range) • Estimate E with first guess: OK result • Re-estimate f -1 from (Z,X) pairs... Just linear least-squares!!! • Iterate until converged Results from Paper

  4. Results Debevec HRD Examples • http://gl.ict.usc.edu/Data/ HighResProbes/hdrvr/uffizi.html “Floating Point” HDRI Hugin • http://hugin.sourceforge.net/ Note: the DISPLAY shows everything above 1.0 as white Float/2 Fixed/2 HDR and FP subtly di fg erent

  5. HDR Motion Real vs. Fake Out-of-focus Blur

Recommend


More recommend