Motion picture high dynamic range imaging

Abstract

A process is provided for providing high contrast motion picture imagery. The camera system includes a physical architecture that enables multiple sensors to simultaneously acquire a visual scene without parallax or dissimilar exposures. The system architecture enables current off the shelf components to be used to develop the architecture or a specific sensor array to be developed. The architecture includes optical components that split an incident light ray into multiple beams of varying intensity. Each respective split light ray is then attenuated further with neutral density filters for the wavelengths to be observed before the light ray impacts the respective pixel sights. A process has been developed to use the camera system architecture to stochastically recombine the acquired energy and estimate what the original energy was before attenuation. The process involves the development of an optimal non recursive weighting function that can be used and constructed into a lookup table. The weighting function is used to minimize the error associated with symmetrically distributed error sources. A second filtering routine then takes place to remove the nonsymmetrical sources of error such as intraocular flare. This is accomplished by estimating the symmetric errors prior to acquisition and detecting when a sensors radiant exitance estimate does not fall within this symmetrical error bounds. The final reconstructed HDR image is formed once a pool of cameras is determined and a weighting average of the viable guesses is used to estimate the actual scene radiant exitance, or in a non-unitary sense, a photometric radiance. The HDR image can be saved as a high bit depth image or tone-mapped either by the method introduced or by other means and stored to memory via convention 8-bit encoding methods for storage.