Published Paper Details:

This report presents an in-depth exploration of image segmentation techniques, focusing on Gaussian Mixture Models (GMM) and their comparison with K-Means clustering and Otsu's Thresholding. Image segmentation is a crucial process in computer vision, widely used in applications such as medical imaging, object detection, and autonomous navigation. Gaussian Mixture Models (GMM) provide a probabilistic approach to segmentation, effectively modelling complex distributions within an image. By utilizing multiple Gaussian distributions, GMM can distinguish different regions based on their pixel intensity distributions, making it highly effective for images with overlapping clusters or multi-modal intensity distributions. However, this method comes with a computational cost, as it requires iterative Expectation-Maximization (EM) optimization, which can be slower compared to other clustering methods. This study highlights the flexibility of GMM in handling complex distributions, its trade-off with computational efficiency, the practicality of K-Means for quick clustering, and the simplicity of Otsu's method for threshold-based segmentation. Experimental results demonstrate the effectiveness of these techniques in segmenting images, providing insights into their practical applications, strengths, and limitations. The comparative analysis offers guidance on selecting the appropriate segmentation technique based on specific application requirements.