Constrained Gradient Integration for Improved Image Contrast

Ali Alsam and Hans Jakob Rivertz

Keywords

height from gradient, high dynamic range compression, color to gray conversion.

Abstract

The height from gradient algorithm proposed by Frankot and Chellappa has been employed as a solution to the gra- dient integration problem in both HDR compression and image to gray conversion. In both applications, the algo- rithm results in images that lack local contrast. In this pa- per we present a mathematical modification to the algo- rithm that is shown to maintain the speed advantage while clearly enhancing the local image contrast. Our modifi- cation is based on the observation that the result of HDR compression and color to gray conversion need to be close to the original image luminance. Thus by constraining the output of the height from gradient algorithm to be close to a gamma corrected luminance we are able to compress the large gradients in the low frequency part of the image and enhance the high frequencies. When applied to the prob- lems of HDR compression and image to gray conversion the algorithm results in clear enhancements of the final image contrast over the original Frankot and Chellappa solution.

Important Links:

Go Back