Digital Image Classification Techniques: A Comprehensive Review

Digital Image Classification Techniques: A Comprehensive Review

Utkarsh Shrivastav (Lovely Professional University, India) and Sanjay Kumar Singh (Lovely Professional University, India)
Copyright: © 2019 |Pages: 26
DOI: 10.4018/978-1-5225-9096-5.ch009
OnDemand PDF Download:
No Current Special Offers


Image classification is a technique to categorize an image in to given classes on the basis of hidden characteristics or features extracted using image processing. With rapidly growing technology, the size of images is growing. Different categories of images may contain different types of hidden information such as x-ray, CT scan, MRI, pathologies images, remote sensing images, satellite images, and natural scene image captured via digital cameras. In this chapter, the authors have surveyed various articles and books and summarized image classification techniques. There are supervised techniques like KNN and SVM, which classify an image into given classes and unsupervised techniques like K-means and ISODATA for classifying image into a group of clusters. For big images, deep learning networks can be employed that are fast and efficient and also compute hidden features automatically.
Chapter Preview

Introduction To Digital Image

An image is a copy of a subject that is visibly similar to that object. An image a 2D representation of any certain object. There can be many examples of an image, people go to some places or cherish some moments and take pictures. So it can be concluded that images show similarity or resemblances from the reality. In current scenario mostly digital images are being captured and used. Digital images is a representation of an image in form of numbers, normally binary, of a two-dimensional image. These images are generally captured with the help of a digital camera and then they are transferred to any system having processing probabilities for further processing of image. The processing on any digital image is known as digital image processing. The processing may consist of different functionalities to achieve some particular characteristics of an image like image enhancement, color transformation, spatial transformation. (Gonzalez & Woods,2008)

Image Enhancement

In digital image, it may happen that there may be some sort noise or disturbance that is included in the image. It may be due to the poor camera quality, camera movement, bad light conditions. So to remove that, people often use image processing so that those minor yet affecting impurities can be removed or minimized. (Khamparia & Pandey, 2015) The image enhancement functionalities may include functions like: enhancing the contrast, brightness, sharpness of an image, removing red light affect on any image. Figure 1 is showing before and after effects of image processing. figure 1(a) shows, the image is dark also contrast is high and figure 1(b) shows that after applying image processing image quality has improved.

Figure 1.

Image Enhancement


Color Transformation

In the past, there were Mechanical cameras in which the negative images were produced and it was impossible to make changes in the color or type of the image. (Khamparia & Pandey, 2017) But now, digital images gives freedom to change image of one type into image of another type for example a color image can be converted into a black and white image. also any video can be converted into any format. In the Indian film industries there are movies like “Mughal-e-azam” which was earlier released in black and white form but then it was re-released in color format. (Unnithan,2004)

Figure 2 shows color transformation in a sample image, which was RGB earlier and then converted to grayscale image.

Figure 2.

Color Transformation


Spatial Transformation

Sometimes, it may require to bring structural transformation of image like crop, resize, rotate, scale etc. Image processing allows someone to perform such transformations. One can perform such operations with various tools.

Image processing addresses the problems like resolution fixing, focusing, color saturation fixing, hue fixing etc. As fast as technology is growing the use and demand of image processing is also growing. Like for criminal detection or border security the use of cctv cameras are used. Similarly, for weather forecasting (Deeprasertkul & Praikan,2016) and even in medical uses (Birare & Chakkawar,2018) the use of image processing is growing day by day.

Further more this chapter has been divided into 4 sections, section 2 discusses about the various types of images. Section 3 describes about the classification and various techniques used for image classification purpose. Section 4 talks about deep learning concepts and some of the very popular deep learning networks like convolutional neural network and autoencoders.

Complete Chapter List

Search this Book: