<![CDATA[The growing needs for efficient image data compression have been driven by rapid technological advancement. This study evaluates the effectiveness of the K-Means algorithm for digital image compression through dominant color palette extraction and subsequent color quantization for image reconstruction. The research investigates how the number of clusters (K) affects both compression ratio and Peak Signal to Noise Ratio (PSNR). An experimental approach was implemented, compressing 24-bit RGB images at various resolutions (VGA, SVGA, HD, FHD) using different cluster quantities (8, 16, 32, 64, 96, 128). Through descriptive and correlation analyses, relationships between cluster numbers, compression ratio, Mean Squared Error (MSE), PSNR values, and processing time were examined. Results demonstrate that the K-Means algorithm achieves effective image compression, with an average compression ratio of 67%, MSE of 0.00077, PSNR of 81.43 dB, and processing time of 0.73 seconds. Compression quality was strongly influenced by cluster quantity, with both PSNR values and compression ratios improving as cluster numbers increased. The research determined that 96 clusters represents the optimal configuration, delivering high-quality compression with reasonable computational efficiency]]>
Copyrights © 2025
