Detecting motion in high resolution video images

• Main Authors: U.S. YADHUNANDAN, GURUMUTHY SWAMINATHAN, BEN A MILLER
• Format: Patent
• Language: English
• Published: 28.05.2014
• Subjects: CALCULATING; COMPUTING; COUNTING; ELECTRIC COMMUNICATION TECHNIQUE; ELECTRICITY; IMAGE DATA PROCESSING OR GENERATION, IN GENERAL; PHYSICS; PICTORIAL COMMUNICATION, e.g. TELEVISION

A method for detecting motion in a high-resolution video involves the steps of dividing 102 a video frame into a number of blocks, determining 104 whether motion is present in each of the blocks and, for those blocks in which motion is determined to be present, identifying 108 pixels in which motion is present. Histogram techniques may be used to detect motion in the blocks. The blocks in which motion is identified may be marked 106 as seed blocks. A subsequent image frame may be subdivided likewise 110 and a determination made to identify motion 112 in subsequent image blocks corresponding to the seed blocks and its neighbouring blocks from the first image frame. If motion is determined in the subsequent image frame blocks, motion detection is performed to identify 114 pixels in the subsequent image frame in which motion is present. The pixels in which motion is present can be identified, for example, using a fast pixel based motion detection algorithm such as a kernel density estimation (KDE) or Gaussian mixture model (GMM) based motion detection algorithm. The method allows analysis of a high-resolution image frames to identify motion without necessarily performing pixel motion detection in all image regions which would otherwise be too computationally expensive to support real-time motion detection.