Noticeable Difference

Just Noticeable Difference (JND) research focuses on determining the minimum perceptible change in a stimulus, primarily visual, that a human can detect. Current research emphasizes developing accurate JND models across various distortion types (e.g., compression, adversarial attacks) and incorporating human visual system (HVS) characteristics, often using deep learning architectures like convolutional neural networks to predict JND maps. This work is crucial for optimizing image and video compression, improving face recognition systems, and enhancing privacy-preserving techniques by enabling the generation of perceptually indistinguishable yet functionally altered images.