Comparametric image compositing: Computationally efficient high dynamic range imaging
Mir Adnan Ali, Steve Mann · 2012
We propose a novel computational method for compositing low-dynamic-range (LDR) images into an high-dynamic-range (HDR) image by the use of a comparametric camera response function (CCRF), which is the response of a virtual HDR camera to multiple inputs. We demonstrate the use of this method with a simple probabilistic joint estimation model, that accounts for Gaussian noise, using iterative non-linear optimization to compute the CCRF. We achieve a speedup of ≈2500×, relative to direct calculation using the probabilistic model. This method can be implemented as a multidimensional lookup table, and enables realtime HDR video with any camera response function model, and any compositing algorithm based on pixel value and exposure.