Publication: Nonlinear exposure intensity-based histogram equalization for non-uniform illumination image
Loading...
Date
2022-09-01
Authors
Saad, Nor Hidayah
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Non-uniform illumination image consists of different illumination regions. Applying the same contrast enhancement concept to the whole image can over or under enhance the image. Thus, different enhancement concepts should be applied to different illumination regions, necessitating region determination. Almost all existing region determination methods only consider intensity to determine the regions, so they can only detect two different regions, dark and bright, which does not represent the real exposure condition. To solve this, Local Neighbourhood Exposure Region Determi nation (LNERD) is proposed by considering the intensity, entropy and contrast which can better represent the details in the image. The three attributes are combined with a rule-based method for identifying illumination regions before enhancement process is applied. Due to over-enhancement, the existing Histogram Equalization (HE)-based methods produce washed-out effects and unnatural appearance, limiting the ability to achieve illumination uniformity of an image. Therefore, this study proposes a modified HE method named as Nonlinear Exposure Intensity Modification Histogram Equaliza tion (NEIMHE). The proposed NEIMHE method divides exposure regions into five sub-regions and modifies each sub-histogram region’s by adding a nonlinear weight to its cumulative density function (CDF). The modified HE equations provide intensity expansion and mapping directions for under- and overexposed regions. Using 600 non-uniform illuminated sample images from three databases, the proposed LNERD method detects over-exposed, well-exposed, and under-exposed regions more accurately than existing methods. The proposed NEIMHE method improves image unifor
mity, detail, and naturalness by achieving the highest scores in Discrete Entropy (DE),Measure of Enhancement (EME), and Peak Signal to Noise Ratio (PSNR) and secondbest in Absolute Mean Brightness Error (AMBE) and Lightness Order Error (LOE).Those findings prove that NEIMHE can improve non-uniform illumination images’exposure.