
		<paper>
			<loc>https://jjcit.org/paper/221</loc>
			<title>ILLUMINATION ENHANCEMENT OF NIGHTTIME IMAGES USING A REGULATED SINGLE SCALE RETINEX ALGORITHM</title>
			<doi>10.5455/jjcit.71-1705156218</doi>
			<authors>Ola A. Basheer,Zohair Al-Ameen</authors>
			<keywords>Nighttime,Image enhancement,Single-scale retinex,Statistical methods</keywords>
			<citation>4</citation>
			<views>3680</views>
			<downloads>471</downloads>
			<received_date>14-Jan.-2024</received_date>
			<revised_date>  11-Mar.-2024</revised_date>
			<accepted_date>  13-Mar.-2024</accepted_date>
			<abstract>Nowadays, people are active during the nighttime and take many photos to record their activities. Due to the low-
light  nature  of  the  environment at  nighttime,  captured  images  tend  to  appear with dimmed and imbalanced 
illumination, limited contrast, covert noise and diminished colors. Thus, this paper presents a practical algorithm 
to improve the illumination of nighttime images based on the single-scale retinex model, image processing methods 
and certain statistical functions. The developed algorithm initiates by converting the image from the RGB into the 
HSV model. Then, it enhances only the value (V) channel while preserving the H and S channels. Next, estimating 
the illumination version of the image and calculating the logarithms of both the illumination and original image 
are performed. Afterward, a  logarithmic subtraction  occurs and a  modified cumulative  distribution  function  of 
Gumble  probability  is  applied and the  result  is  further enhanced using  a  logarithmic  transform  method. These 
operations produce  the  processed V channel and a  conversion  to  the RGB  format occurs  to  generate  the  final 
output.  The  proposed  algorithm  is experimented with  by using two  datasets,  compared to ten different 
contemporary  algorithms and outcomes  are evaluated via three sophisticated metrics. Based  on  the attained 
results, promising performances by the developed algorithm have been recorded, surpassing the performance of 
many existing algorithms in various objective, subjective and runtime terms.</abstract>
		</paper>


