Recursive Median and Mean Partitioned One-to-One Gray Level Mapping Transformations for Image Enhancement

This paper presents two novel recursive partitioned one-to-one gray level mapping (RPOGM) algorithms, viz., recursive median partitioned one-to-one gray level mapping (RMDPOGM) and recursive mean partitioned one-to-one gray level mapping (RMPOGM). The proposed RPOGM methods serve multiple objectives...

Full description

Saved in:
Bibliographic Details
Published in:Circuits, systems, and signal processing systems, and signal processing, 2019-07, Vol.38 (7), p.3227-3250
Main Authors: Reddy, M. Eswar, Reddy, G. Ramachandra
Format: Article
Language:English
Subjects:
Circuits and Systems
Electrical Engineering
Electronics and Microelectronics
Engineering
Equalization
Histograms
Image enhancement
Instrumentation
Mapping
Recursive methods
Signal,Image and Speech Processing
State of the art
Transformations
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:This paper presents two novel recursive partitioned one-to-one gray level mapping (RPOGM) algorithms, viz., recursive median partitioned one-to-one gray level mapping (RMDPOGM) and recursive mean partitioned one-to-one gray level mapping (RMPOGM). The proposed RPOGM methods serve multiple objectives and address the issues such as (i) intensity saturation, (ii) intensity compression and (iii) ensure uniform degree of enhancement of all gray levels and thus result in overall enhancement of the processed image. In RMPOGM, image/histogram is partitioned recursively, (recursion level restricted to two, resulting in four sub-histograms) based on mean. RMDPOGM is similar to RMPOGM except histogram partitioning is done based on median. In RPOGM methods, image-dependent weights for each sub-histogram are calculated separately. Later, these weights are used for transformation. Finally, all the transformed sub-images are combined to get the processed image. As the images processed by these methods are not having any loss of details, it results in retaining the structural details of the objects and hence preserves fine contours even after enhancement. This results in low gradient magnitude similarity deviation (GMSD) between the processed image and input image. Experimental results show the superiority of the proposed methods over the state-of-the-art histogram equalization methods in terms of preserving entropy, preserving mean brightness and having low GMSD.
ISSN:0278-081X
1531-5878
DOI:10.1007/s00034-018-1013-3