Method and system for real-time and offline de-identification of facial regions from regular and occluded color video streams obtained during diagnostic medical procedures

Computer-implemented method, comprising:

• localizing, by a device operatively coupled to a processor, a face depicted in a frame of a video stream using a first process;
• anonymizing, by the device, pixels in the frame that correspond to the face based on the localizing;
• determining, by the device, a structural similarity index that represents a measure of similarity between the frame and a subsequent frame of the video stream;
• and
• tracking, by the device, the face in the subsequent frame of the video stream using a second process different from the first process based on the structural similarity index satisfying a threshold.

A non-transitory computer-readable medium for facilitating automated facial anonymization in video streams, the non-transitory computer-readable medium comprising instructions, that when executed by a processor to cause the processor to:

• localize a face depicted in a frame of a video stream using a first process;
• anonymize pixels in the frame that correspond to the face based on localization of the face using the first process;
• determine a measure of similarity between the frame and a subsequent frame of the video stream;
• and
• track the face in the subsequent frame of the video stream using a second process different from the first process based on the measure of similarity satisfying a threshold.