This paper presents an enhanced image-encryption technique based on the Advanced Encryption Standard (AES). While AES ensures strong security for text and binary data, direct application to images leaves vulnerabilities due to their inherent redundancy and spatial correlations. To address these weaknesses, we introduce a multi-stage preprocessing and key-management framework. First, images undergo frequency-domain transformation via Discrete Wavelet Transform (DWT) or Discrete Cosine Transform (DCT) to reduce redundancy and highlight detail components. Next, chaotic-map-based pixel shuffling (using Logistic and Arnold Cat maps) disrupts spatial relationships. Dynamic session keys—generated from biometric inputs and time-stamps—replace fixed AES keys to thwart key-recovery attacks. Finally, optional parallel processing accelerates the AES encryption/decryption pipeline for large or real-time datasets. Experimental results on standard test images demonstrate significant improvements: entropy increased from 7.3 to up to 7.95 (ideal = 8), NPCR rose from 88% to over 99%, UACI from 25% to over 34%, and correlation coefficients dropped near zero. These enhancements yield stronger resistance to statistical and differential attacks with only modest additional computational overhead.