Spatio-Temporal Retinex-Inspired Envelope with Stochastic Sampling: A Framework for Spatial Color Algorithms

摘要

The authors present a new framework for algorithms for a wide range of image enhancement and reproduction applications, named stress-Spatio-Temporal Retinex-inspired Envelope with Stochastic Sampling. The algorithms work by recalculating each pixel using envelopes for local upper and lower bounds in the image. The envelopes are obtained sampling neighbor pixels and can be interpreted as local reference maximum and minimum. This approach derives from a computational simplification of previous spatial color algorithms like Retinex or ace. With the proposed method, various tasks such as local contrast stretching, automatic color correction, high dynamic range image rendering, spatial color gamut mapping, and color to grayscale conversion can be performed with good results. The algorithm exhibits behaviors in line with some aspects of the human visual system, e.g., simultaneous contrast.