Videos and Presentations

Tone Mapping Optimization

When comparing the output capabilities of standard display systems and the perceptual capabilities of the human visual system the greatest disparity exists in their dynamic ranges. The dyanmic range of most displays is well below the in-scene dynamic range of the visual system, while high dynamic range (HDR) images can have dynamic ranges that approach the full dynamic range limits of human vision. Tone mapping algorithms, HDR image compression algorithms, compress the dynamic range of HDR images to a lower their dynamic range, while keeping the details and features visible in the low dynamic range (LDR) image. These algorithms do this by taking advantage of the fact that the human visual system is not sensitive to absolute luminance values, it is more sensitive to the logarithm of the luminance. Most tone mapping algorithms are designed with the goal of creating aesthetically pleasing LDR images or LDR images that are perceptually similiar to the original HDR image.

This project aims to first design an automated tone mapping algorithm that improves the amount of information or detail in the LDR image. In order to find the optimal tone mapping operator that shows the maximum amount of visible information in the LDR image an image quality metric that quantifies the amount of detail in a LDR image will be developed. Then using the quality metric as an measure of error and a tone mapping model that can emulates most tone mapping methods an optimal tone mapping operator that maximizes the amount of detail in a compressed image can be found.