IMAGE SEGMENTATION USING GAUSSIAN MIXTURE MODELS

• Main Author: Iffadathul Faddilla, Nor
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/44308
• Institution: Institut Teknologi Bandung

In this digitalization era, many applications want to represent images with as little memory as possible. In addition, data collection and storage technology are increasing rapidly. Image is a multimedia component that has an important role in a form of information. A lot of information and meaning are contained there. The larger the size of the image, the greater the memory and speed necessary to access or transmit the image. One of the problems in image processing is how to obtain information from the image as original sound as possible. In addition, many of the images that contain duplication of data because of the intensity of the pixels is equal to its neighbors or parts of the same image with other parts which resulted in wasteful of memory. In this final task, we use grouping or clustering from pixel color images that have the same features or called image segmentation. The goal of image segmentation is to simplify or change the representation of an image into something that is more meaningful and easier to analyze. Image pixels must be grouped into classes according some criterion, called cluster. The density of cluster is modeled by Gaussian Mixture Models. To estimate the parameters in the proposed model, the Expectation-Maximization (EM) is applied. For varying number of clusters we obtained images with different information. To determine the maximum value of cluster optimal, we use of Calinski Harabasz Criterion