Content-Based Video Preprocessing for Remote Monitoring of Neurosurgery

Transmitting high-quality video via the Internet is a challenging problem in distributed diagnosis system. We propose a content-based video data preprocessing scheme, which is adaptive to the video data acquired during intraoperative monitoring and can cascade with essentially any video codec. IOM video is first decomposed into temporal wavelet subband frames and motion compensation is incorporated into these transforms to exploit inter-frame redundancy efficiently. Highpass subband frames are then adaptively weighted according to an importance map, which specifies the importance of the video contents for clinical observation. This map has higher value near the surgical site and lower value at the surrounding area. The surgical site can be located automatically by a surgical tool tracking system