Lorenz Curve-Based Entropy Thresholding on Circular Histogram

Chao Kang,Jiulun Fan,Chengmao Wu

doi:10.1109/access.2020.2964335

Chao Kang, Jiulun Fan + Show 1 more

Open Access

https://doi.org/10.1109/access.2020.2964335

Copy DOI

Abstract

Circular histogram thresholding is a new threshold selection method in color image segmentation. However, the method of the existing circular histogram thresholding based on the Otsu Criteria lacks the generality of using the circular histogram. In order to improve the effectiveness and reduce the complexity of thresholding on circular histogram, this paper firstly introduces the Lorenz curve into circular histogram. Then the circular histogram is expanded into the linearized histogram in clockwise or anti-clockwise direction by the optimal index of the Lorenz curve. In the end, the entropy thresholding of the linearized circular histogram is adopted to choose the optimal threshold to obtain the object of color images. Many experimental results show that the proposed method has better effectiveness and adaptability than the existing circular thresholding utilizing Otsu Criteria.

Highlights

Target acquisition of color image is an important problem in image analysis, understanding and computer vision due to the color images can provide more information than graylevel images
In order to reduce the complexity of entropy-based circular thresholding and make it more adaptable to be extended to multi-thresholding cases, we aim to break the circular histogram into the linearized one and obtain the optimal threshold on that
EXPERIMENT RESULTS AND ANALYSIS In order to evaluate the performance of our proposed methods, all the experiments are simulated on PC with Matlab (2014 version) on H component

Summary

Introduction

Target acquisition of color image is an important problem in image analysis, understanding and computer vision due to the color images can provide more information than graylevel images. Among all color spaces, RGB is not suitable for the color image segmentation due to the high correlation among the three components [3]. HSI color space [2], [4] has been used more frequently due to three highly irrelevant components: hue (H), saturation (S) and intensity (I). Many researchers had used the HSI color space in the color images’ segmentation [4]–[7] due to this model which can represent hue information and has a more natural correspondence to human vision than the RGB color model.

Objectives

Results

Conclusion