Digital Image Enhancement in Matlab: An Overview on Histogram Equalization and Specification

This paper has two major parts. In the first part histogram equalization for the image enhancement was implemented without using the built-in function in MATLAB. Here, at first, a color image of a rat was chosen and the image was transformed into a grayscale image. After this conversion, histogram equalization was implemented on the grayscale image. Later on, in the same image for each RGB channel, histogram equalization was implemented to observe the effect of histogram equalization on each channel. In the end, the histogram equalization was implemented to this specific color image of a rat. In the second part, for the grayscale image in part 1, the desired histogram of another colored image of a rat was introduced and histogram specification was implemented on the original colored image.

I. Introduction

In the modern age, image enhancement technique becomes a vital tool to facilitate with the improvement in image quality in various sectors like identifying anything in images as well as medical imaging [1], [2], computational photography, forensic analysis [3], and pattern recognition [4] in machine vision applications. Main motive of this technique is to make the images discernible by correcting the color hue and Brightness imbalance [5] as well as contrast adjustment [6]. As the background of an image may hide structural information of an image [7], the technique to prolong the image temperament, enhances the foreground information, while retaining the background information and thus increases the overall contrast of an image [8], [9]. Several algorithmic techniques such as Artificial Neural Network [10], Convolutional neural Network [11], and K-nearest Neighbors [12] can also be applied in image processing techniques such as segmentation, thresholding and filtering. Though there are several image enhancement techniques has been developed over the past decades, the histogram based image enhancement techniques specifically; 1) Histogram equalization, 2) Histogram specification are utilized vastly for their high efficiency and simplicity of algorithm [5].