Improved image enhancement method for non-uniform illumination and low ontrast images using bihistogram modification approach
Loading...
Date
2016-07-01
Authors
Kong Teck Long
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
In certain situations, low contrast and non-uniform illuminated images would be
captured. These images are considered as challenge in the field of computer vision and pattern recognition. The conventional techniques that are commonly used to solve this problem have some limitations. Some of these methods require manual parameters tuning. Besides that, some of the methods are only focused on one or two aspects of noise reduction, contrast enhancement, non-uniform illumination enhancement and detail preservation. Hence, this study proposes a new method which is Bi-histogram Modification for Illumination Correction (BHMIC). The proposed BHMIC will first distinguish the bright and dark regions of an image. Then, it is followed by enhancing the contrast and illumination condition of the image. At the same time, filtering process is employed in order to remove the details of the image (i.e. edges) and noises. It is done to preserve the details and avoid the amplification of noises. The proposed method applies the illumination and reflectance assumptions to separate the dark and bright regions of the image. Then, these regions are enhanced using derived dark and bright enhancers separately. The
modified clipped histogram equalization is then applied for contrast enhancement
purpose. Finally, the details of the image are added to the illumination corrected and contrast enhanced image as an output image. Qualitative analysis shows that the proposed BHMIC has good performance in detail preservation, contrast enhancement and illumination condition enhancement without significantly amplifying unwanted noise. The quantitative analysis shows that the proposed BHMIC is 38% to 42.9%, 0.8% to 4.7% and 0.7% to 2.3% better than other tested methods in EME, NIQE and entropy, respectively. The promising results suggest that the proposed BHMIC could probably be used in pre-processing of face images for face recognition, medical images for easier disease symptoms diagnosis and photography images for personal usage.