Extrapolation of Saturated Diffraction Spikes in Photographs Containing Light Sources

Abstract

The observed luminance of most light sources is many orders of magnitude higher than the luminance of surrounding objects and the background. With the dynamic range of single photographs limited to 8 to 10 bits, both dark and bright values are radically coerced to the limits. This problem is usually circumvented with the use of High-Dynamic-Range (HDR) imaging, assembling multiple photographs of the same scene made at different exposures. But in some situations, or due to equipment limitations, HDR imaging might not be possible. This research is aimed at the extrapolation of luminance peaks within oversaturated Low-Dynamic-Range (LDR) images. The proposed method of extrapolation relies on the identification and analysis of Fraunhofer diffraction patterns created by the aperture. The algorithm is tested on a set of HDR images containing one or two lamps. These images are converted to LDR at a custom saturation cap, then extrapolated to restore the original peaks with relative success.