Introduction to image segmentation in Computer Vision
Image segmentation is a subset of computer vision where Computer Vision is a subset of Artificial Intelligence. Computer vision system is build based on block of various stages. Today world many of the application using image segmentation technology such as traffic control system, self-driving cars, locating objects in satellite images. The process of distinguishing objects from an image called image segmentation. Commonly there are two type of segmentation technique using on image processing in computer vision.
1. Semantic Segmentation
Labeling the pixels in the image on a specific common class.it means every pixel belongs to a particular class. Each pixel with a specific class is addressed in the same color (image 1).
2. Instance Segmentation
The process of instance segmentation is distinguishing each object instance for every known object on the image.it also assigned a particular class to each pixel of the image. Commonly this segmentation used for counting number of object on the image (image 2).
The image pixels information classify into two categories based on the features (value).1. Dynamic Information (Foreground): the moving or nearest to viewers objects on the image called foreground object. The object pixels considered as foreground pixels.
2. Static Information (Background) : Non-Moving or far to viewers objects on the image called background object. Commonly the object pixels knows as background pixels.
The image segmentation process ability to handing various difficult scenario such as illumination changes, camera motion, camouflage effect, shadow etc. In the process of image Segmentation, planning at a certain rate can improve the clarity of image pixels, and the image quality can be significantly improved. It is important to establish a proper connection for the segmented image, and on this basis, it cannot be accessed and repeated. At the same time, it is important to ensure that the segmented image is highly consistent and the image will not change. Image segmentation and feature extraction transform the original image into abstract form for advanced image analysis and understanding, which lays a good foundation for better application of image segmentation technology.
Also some methods are using on image segmentation technology. Here such methods are briefly explain below,
Method 01: Threshold Segmentation Method
Threshold segmentation methods is simplest and widely used method on image segmentation. The method divide the image pixels by determine the threshold value based on pixels intensity level. Same pixels objects are segmented uniformly, it increase the accuracy level of segmentation process.
The threshold process, comparing every pixel value of the image to a determined threshold value. These comparison segment all the input image into two category
1. Pixel’s value lower than threshold
2. Pixel’s value greater than threshold
Input image f, output image g and T is a threshold value. The Image pixels of the object refer as g(i,j) = 1(foreground) and the image pixels of the background refer as g(i,j) = 0. The main function of the methods convert the input image as a binary image.
Method 02. Clustering Segmentation method
Segment the image into clusters having pixels with similar characteristic. Displays different regions of the image according to the image set, and the same region and the same Chroma are well fused between the color image and the gray image, and then the image segmentation is realized. The main goal of the method, perform effective aggregation analysis and real-time segmentation.
Method 03. Edge based segmentation method
The principal of edge based segmentation method, detect the most edge of the image and create the independent part from the image based on edge pixel value. Usually the easiest place to detect a part of an image is the edge of the image. The most intense change of image information takes place at the edges of the image. find the edge of the image using the sobel operator function. Sobel operator is found according to the gray difference between the upper and right adjacent regions, and reaches the maximum value at the edge. Edge based segmentation method provide more efficient information.