VRL an eye

High Dynamic Range imaging and compression

People

Dmitry V. Fedorov, Baris Sumengen, B.S. Manjunath

Objective

Dynamic range of modern aquisition systems is lower than the one of human beings. It's usual that certain objects appear too bright or even saturated and other objects are not visible because of darkness depending on the camera settings (see Fig. 1). Our framework for image processing using local information provides a solution combining images taken at different exposures and compressing them to the desired Dynamic Range.

Img. 1a) photo at EV-7 Img. 1b) photo at EV 0 Img. 1c) photo at EV+5

Example 1

The first example (a) shows Tone mapping operator of Ward Larson et al. (1998) that preserves well the relative contrasts. The example (c) of Gradient Domain HDRC by Raanan Fattal et al. (2002) preserves well local information. The images of Stanford Memorial Church are taken from Paul Debevec web page on Radiance.

Img. 2a) Tone mapping operator of Ward Larson et al. (1998) Img. 2b) Our result Img. 2c) Gradient Domain by Raanan Fattal et al. (2002)

Example 2

The second example shows Tone mapping operator (a) of Ward Larson et al. (1998) that preserves well the relative contrasts. The example (c) of Gradient Domain HDRC by Raanan Fattal et al. (2002) preserves well local information. The Belgium House example taken from Dani Lischinski web page on Gradient Domain High Dynamic Range Compression.

Img. 2a) Tone mapping operator of Ward Larson et al. (1998) Img. 2b) Our result Img. 2c) Gradient Domain by Raanan Fattal et al. (2002)

Example 3

This example shows of Gradient Domain HDRC (a) by Raanan Fattal et al. (2002) preserves well local information. The chair example taken from HDR images from the CAVE (Columbia Automated Vision Envirenment) Lab.

Img. 2a) Gradient Domain by Raanan Fattal et al. (2002) Img. 2b) Our result

Example 4

This example shows of Gradient Domain HDRC (a) by Raanan Fattal et al. (2002) preserves well local information. The chair example taken from HDR images from the CAVE (Columbia Automated Vision Envirenment) Lab.

Img. 2a) Gradient Domain by Raanan Fattal et al. (2002) Img. 2b) Our result

Acknowledgements

This project is supported by the NSF Information Technology Research grant #0331697.