Image Encryption using Color Space Neural Network Transformation and Chaotic Pixel Perturbation
摘要
This paper presents a novel image encryption method combining neural networks, color space transformation, and chaotic systems. Unlike traditional methods, it enhances the dimensionality of complexity by introducing key- and input-dependent security. It integrates dual-stage pixel scrambling, memory-based bidirectional substitution, and color transformation neural networks. The proposed method yields low adjacent pixel correlation (0.0002) and near-ideal entropy (7.9990), with minimal computational load.