Technical Program

Paper Detail

Paper:	IMDSP-P17.6
Session:	Image Quality Assessment and Enhancement
Time:	Friday, May 19, 16:30 - 18:30
Presentation:	Poster
Topic:	Image and Multidimensional Signal Processing: Restoration and Enhancement
Title:	AUTOMATIC DIGITAL IMAGE ENHANCEMENT FOR DARK PICTURES
Authors:	Jie Zhao, Shawmin Lei, Sharp Labs of America, United States
Abstract:	An automatic method is presented to enhance the visibility in dark areas of an image while preserving its natural look. This method consists of three major parts: classification, global adjustment and local adjustment. Firstly, an image is classified into one of several types for different global/local processing or no processing. The method adaptively maps a luminance value into a value based on a piecewise linear mapping curve in order to increase contrast in dark areas. The global and local mapping curves are based on histogram equalization with modifications. The local enhancement is non-overlapped block-based and is applied only when necessary. To avoid blocking artifacts generated by block-based local adjustment, the luminance value of each pixel is adjusted according to an interpolated mapping curve derived from the block mapping curves of nearby blocks. The combination of global and local enhancements achieves good overall enhancement with local adaptability without excessive complexity.