This project, by Kaifan Zheng and Noshin Chowdhury, introduces an efficient image segmentation method using Principal Component Analysis (PCA) to optimize Gaussian Mixture Models (GMM) and Expectation-Maximization (EM) Algorithm.
- PCA: A technique for data structure exploration and dimensionality reduction.
- K-Means Clustering: A simple clustering algorithm for local optima computation.
- GMM and EM Algorithm: Used for estimating optimal parameters and clustering discrete data.
- The combination of PCA with GMM and EM reduces segmentation time significantly (~25x faster) while maintaining high accuracy.
- The approach is particularly effective for large image datasets.
- Experiments show improved speed and accuracy in image segmentation tasks.
- Test results demonstrate efficient mean color segmentation and accurate color averaging for smoother output images.
Labled Image , k = 7
averaged Image , k = 7
- For detailed mathematical formulations and algorithmic explanations, refer to the cited bibliography in the report.