Expectation-Maximization Algorithm and Image Segmentation

In computer vision, image segmentation problem is to partition a digital image into multiple parts. The goal is to change the representation of the image and make it more meaningful and easier to analyze [11]. In this assignment, we will show how an image segmentation algorithm works in a real application. In the Electronic Field Guide (EFG) project, researchers want to segment the leaf region from an image and extract a set of points from the contour to represent the shape of leaf [2]. A leaf image typically contains a single piece of leaf on a surface with rather uniform pattern. This makes the segmentation problem easier. Figure 1 shows an example of leaf images. A matrix of pixels represents a leaf image. Each pixel is a 3 by 1 vector representing the value of red, green, and blue components respectively. Each value is an integer value between 0 to 255. Instead of working on this 3-dimensional data, we transform a pixel in hue, saturation, and value domain (HSV) [7], and get rid of the hue component. Our task is to group this 2-dimensional data into two clusters: the leaf region and non-leaf region.