Aspects of anonymization of audio-visual medical data are described. Video data comprising a sequence of images is received for monitoring a subject. One or more faces are detected in each image of the sequence of images. A face of the subject is identified amongst the one or more faces. The identified face is tagged as subject face and remaining faces are tagged as bystander faces. The bystander faces are masked. It is determined whether live monitoring is to be performed for the subject. Based on the determination, the subject face is morphed to obtain anonymized video data, wherein the morphing comprises changing landmark features without changing facial expressions.