Contrast Enhancement using Completely
Overlapped Uniformly Decrementing Sub-Block
Histogram Equalization for Less Controlled
Illumination Variation
Shree Devi Ganesan1 and Munir Rabbani2
1Department of Computer
Applications, B.S. Abdur Rahman Crescent Institute of Science and Technology,
India
2Department of
Mathematics, B.S. Abdur Rahman Crescent Institute of Science and Technology,
India
Abstract: Illumination pre-processing is an inevitable step for a real-time automatic face recognition system in solving challenges related to lighting
variation for recognizing the face images. This paper proposes a novel framework namely
Completely Overlapped Uniformly Decrementing Sub-Block Histogram Equalization
(COUDSHE) to normalize or pre-process the illumination deficient images.
COUDSHE is based on the idea that efficiency of the pre-processing technique mainly
depends on the framework for application of the technique on the affected
image. The primary goal of this paper is to bring out a new strategy for
localizing a Global Histogram Equalization (GHE) Technique to help it adapt to
the local light condition of the image. The Mean Squared Error (MSE), Histogram
Flatness Measure, Absolute Mean Brightness Error (AMBE) are the objective
measures used to analysis the efficiency of the technique. Experimental Results
reveal that COUDSHE has better performance on Heavy shadow images and half lit
image than the existing techniques.
Keywords:
Illumination
pre-processing; global histogram equalization; localization; mean squared
error; histogram flatness measure, absolute mean brightness error.