Comparision of Principal Component Analysis, K-means, and Generative Adversarial Network for Compression and reconstruction of medical image

Medical images are essential in healthcare, but their large file sizes present challenges for storage and transmission. This study proposes a framework for compressing color medical images while preserving key visual information. We apply three compression models: Principal Component Analysis (PCA), K-Means clustering, and Generative Adversarial Networks (GANs), and evaluate their performance based on Compression Ratio (CR), Peak Signal-to-Noise Ratio (PSNR), Mean Squared Error (MSE), and Structural Similarity Index (SSIM). K-Means with 100 clusters outperforms the other models, achieving the highest compression (CR of 1.20), with excellent image quality (PSNR of 66.77 dB, SSIM of 1.0000). PCA shows moderate performance with a CR of 2.45, while GAN provides the least effective compression and image quality. These results demonstrate that K-Means offers the best trade-off between compression and image fidelity.