Image segmentation is gauged as an essential stage of representation in image processing. This process segregates a digitized image into various categorized sections. An additional advantage of distinguishing dissimilar objects can be represented within this state of the art. Numerous image segmentation techniques have been proposed by various researchers, which maintained a smooth and easy timely evaluation. In this chapter, an introduction to image processing along with segmentation techniques, computer vision fundamentals, and its applied applications that will be of worth to the image processing and computer vision research communities has been deeply studied. It aims to interpret the role of various clustering-based image segmentation techniques specifically. Use of the proposed chapter if made in real time can project better outcomes in object detection and recognition, which can then later be applied in numerous applications and devices like in robots, automation, medical equipment, etc. for safety, advancement, and betterment of society.
TopIntroduction
The major highlighting view of vision is its inference issues. For instance, one have few measurements, and desires to determine its root cause using a mode. There are numerous vital features that differentiate vision from multiple other inferences glitches; mainly, to identify the object items from the data through which an effective decision making can be conducted. For example, it is quite challenging to identify whether a pixel lies on the dalmation (See Figure 1) just by observing at the pixel. The solution to this issue can be resolved by working over the selection of “interesting” sections. Obtaining this sections is recognized as Segmentation.
Image segmentation is indispensable yet precarious component in image analysis, pattern recognition, low level vision and now optimally contribute in robotics. Besides, being one of the problematic and challenging chores in the field, it evaluates the outcome`s quality for image analysis. Intuitively, image segmenting is termed as a procedure of image segregation into multiple segments such that each segment is homogenous while not unified to any two contiguous segments. A supplementary requisition with image segmentation would be segments correspondence to real identical regions those belongs to objects in the respective scene.
The broadly accepted formal image division is defined as follows:
If HP(o) is considered as a homogeneousness predicate defined on coupled group pixels, then segmentation is defined as a division of set S into multiple associated regions {R1, R2, …., Rn} such that
With R
x ∩ R
y = Փ,
(1)The HP (Ri) is uniform predicate that is true for all regions Ri and HP (Rx U Ry) is false when x! = y and regions Rx and Ry are neighbors.
Moreover, it should be significant to make in consideration that image segmentation problematic issues are one of the psychological perceptions and thus does not exposed to pure analytical solution. This might be the reason behind the proposition of numerous segmentation techniques in the state-of-the-art principally versed on monochrome segmentation and very less survey has been reported over color image segmentation.
By the time passed the attention towards the color image segmentation is attaining due to one of the reasons discussed below:
- •
Reliability: Information provided by the color images is far more reliable in comparison to grayscale images.
- •
Computational Power: The computational power of computers has increased hastily in recent years, even for color image processing.
- •
Database Management: Image database handling, which mainly formed by color images.
- •
Advance Capabilities: There has been an enhancement in sensing proficiencies all intelligent machines.
Numerous image segmentation techniques such as graph based, threshold based, clustering based, morphological based, neural network based and many more are proposed. Each of these technique carries its own pros and cons, therefore the selection of particular technique directly depends on needs from one`s own perspective. Out of these the most effective and optimized technique is clustering based image segmentation. In this chapter, a detailed discussion of various segmentation methods are deeply discussed together with a detail study of several clustering algorithms along with recent updates in the field.
A collective representation of the ideas has been made in this chapter. These methods work over diverse categories of data set: several are envisioned for images, several are envisioned for video series and several to be applied to tokens. (Note: Tokens- it is basically a point that designates the presence of interesting arrangement, such as a dot or edge point (See Figure 1)). Although they all appears to be diversified, there exists a strong similarity among them. Each method tries to attain a compact representation of its respective data sets using various similarity model.