Bilateral Histogram Equalization for Contrast Enhancement

Introduction

Nowadays people capture large amount of images in their daily life, and these images might demand enhancement due to several reasons such as the existence of noise, cloud, low quality of the image capturing devices, etc. (Rahman, 2014). Besides this, images are widely used in many research areas such as atmospheric sciences (Ellrod, 1995), astrophotography (Bedi, 2013), satellite image analysis (Steiniger, 2013), medical image processing (Chronaki, 1997), radar image (Li, & Wang, 1994), texture analysis and synthesis (Efros, 2001; Zhai, 2014), remote sensing (Jensen, 1987), person identification (Kouno, 2012), face image analysis and machine learning (Rahman, 2015; Rahman, 2016).

To enhance the quality of an image, various contrast enhancement techniques have already been proposed such as Histogram Equalization (HE) (Kang, 1977), Brightness Preserving Bi- Histogram Equalization (BBHE) (Kim, 1997), Dualistic Sub-Image Histogram Equalization (DSIHE) (Wang, 1991), Recursively Separated and Weighted Histogram Equalization (RSWHE) (Kim, 2008), Adaptive Gamma Correction with Weighting Distribution (AGCWD) (Huang, 2013), Weighted Adaptive Histogram Equalization (WAHE) (Arici, 2009), Layered Difference Representation (LDR) (Lee, 2013) and Bilateral Histogram Equalization with Pre-processing for Contrast Enhancement (Amil, 2016). Among the existing techniques, HE and its variant such as BBHE and DSIHE are widely used due to its computational simplicity and better performance. HE may over-enhance the whole image because some of the low intensity pixels are transformed at a high rate. To mitigate this problem, several improved versions of HE (i.e., BBHE and DSIHE) are already introduced which divide the histogram into several parts and then apply HE on each part. In this case, the removal of artifacts and desired enhancement depends on the accurate partition of the histogram. BBHE and DSIHE separate the histogram based on the image mean and median, respectively, which can be affected by the outliers. Besides this, the existing image enhancement techniques hardly apply preprocessing techniques, though these processes may help to extract the details of the image.

In this paper, we propose an image enhancement technique to enhance the contrast and extract the image details. Here, the main objective is to bring out the detail information of an image by increasing the contrast of the image without incurring any visual artifact. The contributions of this paper are described as follows.

• 1. This paper proposes Bilateral Histogram Equalization (BHE) where histogram is divided in an accurate manner, and then histogram equalization is performed separately in each part.

• 2. Pre-processing techniques are applied to extract image details and to make image visually smoothing.

• 3. The results of BHE are compared with existing state of the art techniques in both qualitative and quantitative manner.

Literature Review

This section focuses on the pros and cons of several existing image enhancement techniques. Here, at first Histogram Equalization (HE) based techniques are presented, and later some of the gamma correction based techniques are discussed. The details are presented in the followings.

Histogram Equalization (HE) is a widely used technique in image enhancement. HE increases the image contrast by redistributing the pixels of the histogram. Let, X is an input image. Probability density function (PDF) of each pixel can be defined as which is calculated using Equation 1.

Here, and N are the frequency of the pixel and total number of pixels in the image respectively. Cumulative density function is defined by Equation 2. The transformation function is defined by Equation 3. Here, lots of artifacts have been occurred due to the mean-shift problem.