A Beginner's Guide to Multilevel Image Thresholding

Venkatesan Rajinikanth, Nadaradjane Sri Madhava Raja, Nilanjan Dey

• Publisher: CRC Press
• Year: 2020
• ISBN:
  9781000228458
• Edition: 1
• Topic: Computer Science
• Subtopic: Computer Graphics
• Index: Computer Science

A Beginner's Guide to Image Multi-Level Thresholding emphasizes various image thresholding methods that are necessary for image pre-processing and initial level enhancement.

• 
  
  

• Explains basic concepts and the implementation of Image Multi-Level Thresholding (grayscale and RGB images)

• 
  
  

• Presents a detailed evaluation in real-time application, including the need for heuristic algorithm, the choice of objective and threshold function, and the evaluation of the outcome

• 
  
  

• Describes how the image thresholding acts as a pre-processing technique and how the region of interest in a medical image is enhanced with thresholding

• 
  
  

• Illustrates integration of the thresholding technique with bio-inspired algorithms

• 
  
  

• Includes current findings and future directions of image multi-level thresholding and its practical implementation

• 
  
  

• Emphasizes the need for multi-level thresholding with suitable examples

The book is aimed at graduate students and researchers in image processing, electronics engineering, computer sciences and engineering.

1Introduction

1.1Introduction to Image Enhancement

Normally, an image is used to provide the information with the help of picture distribution, and it presents the information in the form of a visual representation method. According to the registration, the image is classified as two-dimensional (2D) and multidimensional (3D), in which the processing procedures existing for the 2D images are quite simple compared to 3D. Furthermore, these images are further classified as conventional image (recorded with gray/RGB scale pixels) and binary image.

Ina variety of domains, images recorded using a chosen modality with a preferred pixel value can be used to deliver meaningful information. In some situations, the information existing in the unprocessed images is hard to understand and hence a number of pre-processing and post-processing procedures are proposed and implemented by the researchers [1, 2]. The implemented image processing schemes can help to improve the state of the raw image with a variety of methodologies, such as contrast-enhancement, edge-detection, noise-removal, filtering, fusion, thresholding, and segmentation [3].

Most of the existing enhancement procedures work well for gray-scale images compared to the RGB-scale images. In literature, a number of image examination procedures are available to pre-process test images and this process will help to convert a raw test image into an acceptable test image [4, 5]. The need for image enhancement and its practical significance is clearly discussed in the upcoming sections for both the grayscale and the RGB scale pictures.

1.2Importance of Image Enhancement

Normally, images recorded using a chosen image modality are referred to as unprocessed images. Based on the need, these raw images may be treated with chosen image conversion or enrichment techniques. Digital images recorded with well-known imaging methodologies may be associated with various problems. Before further assessment, it is essential to improve the information available in the image. In recent times, recorded digital images are processed and stored using digital electronic devices, and in order to ensure eminence, the image needs to be modified to convert the raw image into a processed image. Enhancement procedures such as (i) artifact removal, (ii) filtering, (iii) contrast enrichment, (iv) edge detection, (v) thresholding, and (vi) smoothening are some common procedures that are widely documented in literature are employed to convert the unprocessed image into a processed image. Image enhancement is essential to improve the visibility of the recorded information and to extract the recorded information from the enhanced image is quite easy compared to extracting it from the unprocessed image.

1.3Introduction to Enhancement Techniques

Currently, most of associated imaging systems are processor-controlled systems. Thereby, the images attained from these imaging systems are digital in nature. The quality of the image is judged based on the visibility of the region-of-interest (ROI) and the contrast between the background and ROI. An image recorded with a chosen imaging device is called a raw image. It is to be processed using a chosen image processing technique to convert the raw image into a usable image. This procedure is essential and a complex task when the image ROI has associated unwanted noise and artifact.

Hence, recently a number of image enhancement techniques are proposed and implemented by the researchers, and some of the commonly implemented methods are depicted below.

1.3.1Artifact Removal

This procedure is essential to separate the image into a number of subsections based on a chosen threshold value. In order to group and extract the image pixels into various sections based on the chosen threshold value, the artifact removal technique usually implements a morphological filter along with a clustering approach. This approach is widely implemented in medical image processing applications to remove artifacts available in the 2D slices of magnetic resonance image (MRI) and computed tomography (CT). In order to demonstrate the chosen image enhancement technique (IET), this section considered the 2D CTI of COVID-19 patient obtained from the Radiopaedia database [6]. Figure 1.1 presents the structure of the procedure, and Figure 1.2 presents the results attained on the lung CTI of the COVID-19 image dataset (axial view).

Figure 1.1Implementation of threshold-filter based ROI extraction.

Figure 1.2Experimental outcome with COVID-19 lung CT scan slice.

Threshold filter is widely adopted to pre-process the medical images recorded with imaging modalities, such as MRI and CT. The earlier works on the MRI implemented the threshold-filter approach to separate the brain MRI slices into the skull section and normal brain anatomical regions. The lung CTI examination is also implemented with this technique to extract the lung and the abnormal image parts, such as heart and other body organs. After separating the test image with the chosen technique, the ROI is further considered for the examination process. The earlier works confirms that medical images (MRI as well as CT) pre-processed with this filter helped to attain better results compared to the unprocessed image [7, 8].

Advantages: Threshold filter helps to separate the ROI from the artifact and this process reduces the complexity in the test image to be evaluated.

Limitations: The main limitation of the threshold filter is the selection of finest threshold, which splits the test image into two sections. In most of the cases, the threshold selection is to be done manually with various trials. Trial based technique is a time-consuming process and this technique works only on the grayscale test image.

1.3.2Filtering

In the conventional digital signal processing domain, the filter implemented with a chosen technique and a preferred order is used to allow/block the signal information based on the frequency value. Similar to this operation, the unwanted pixels available in the digital image are r...