Tone Mapping by Interactive Evolution

HDR image source: Mark Fairchild's HDR Photographic Survey

Tone mapping is a computational task of significance in the context of displaying high dynamic range images on low dynamic range devices. While a number of tone mapping algorithms have been proposed and are in common use, there is no single operator that yields optimal results under all conditions. Moreover, obtaining satisfactory mappings often requires the manual tweaking of parameters the influence of which is not always easily understood.

While the quality of a tone mapped image is difficult to assess algorithmically, human subjects typically have no trouble distinguishing well tone mapped images from poorly mapped ones. We thus propose interactive evolutionary algorithms as a tool for tone mapping. Tone mapped images are generated by blending the results obtained from a number of commonly used tone mapping operators in a perceptually uniform colour space. The weights that determine the relative influence of the individual operators together with the operators' parameters form a vector of real-valued variables. An evolution strategy with subjective selection is used to iteratively improve the appearance of the tone mapped images. Importantly, adjustments to the mapping are made in the interactive evolutionary process without a need for the user to understand the influence of the operators' parameters. A user of the system simply needs to pick the most appealing out of a set of automatically generated images. The motivation for blending the images obtained from several tone mapping operators is that mappings that cannot be generated by any one of the individual operators may be achieved. Rather than having to pick the operator most appropriate for the image at hand (and having to accept its limitations), a good mapping that may be outside of the range of any one operator can be found.

Publications

S. B. Chisholm, D. V. Arnold, and S. Brooks
Tone mapping by interactive evolution
Genetic and Evolutionary Computation Conference, Montreal, 2009.

S. B. Chisholm
Tone Mapping by Interactive Evolution
Masters thesis, Dalhousie University, 2009.

Support