Video abstract using signed foreground extraction and fusion

A system for producing a video abstract from video produced by a video camera, the system comprising a central processing unit (CPU) and a graphics processing unit (GPU), the CPU and GPU being configured to perform a method comprising:

• retrieving a frame of the video produced by the camera;
• updating a background frame based at least in part on the retrieved frame;
• performing a video division process to separate static frames from motive frames, the video division process comprising:
• retrieving, if it is determined that the retrieved frame is a static frame, a next frame of the video produced by the camera, updating the background frame, and repeating the video division process;
• and
• extracting, if it is determined that the retrieved frame is a motive frame, foreground data from the retrieved frame after conversion of the retrieved frame to a signed data type;
• determining whether an end of the video produced by the camera has been reached;
• retrieving, if it is determined that the end of the video produced by the camera has not been reached, a next frame of the video produced by the camera, updating the background frame, and repeating the video division process;
• merging, if it is determined that the end of the video produced by the camera has been reached, the foreground data with the updated background data by:
• retrieving updated background frames in time order;
• converting the updated background frames to a signed data type;
• adding the foreground data to corresponding background frames of the converted updated background frames;
• and
• converting the added frames from signed data type to an unsigned character data type to generate a video abstract with the merged foreground and background data.

A method for producing a video abstract from video produced by a surveillance video camera, the method comprising:

• retrieving a frame of the video produced by the camera;
• updating a background frame based at least in part on the retrieved frame;
• performing a video division process to separate static frames from motive frames, the video division process comprising:
• retrieving, if it is determined that the retrieved frame is a static frame, a next frame of the video produced by the camera, updating the background frame, and repeating the video division process;
• and
• extracting, if it is determined that the retrieved frame is a motive frame, foreground data from the retrieved frame after conversion of the retrieved frame to a signed data type;
• determining whether an end of the video produced by the camera has been reached;
• retrieving, if it is determined that the end of the video produced by the camera has not been reached, a next frame of the video produced by the camera, updating the background frame, and repeating the video division process;
• merging, if it is determined that the end of the video produced by the camera has been reached, the foreground data with the updated background data after conversion of the foreground data and the updated background data to the signed data type;
• and
• generating a video abstract with the merged foreground and background data,
• wherein the updating of the background frame comprises:
• converting the retrieved video frame and an initial background frame to a signed data type;
• determining a differential matrix equal to an absolute value of pixel values of the retrieved video frame minus corresponding pixel values of the initial background frame;
• determining whether each pixel value of the differential matrix is less than a defined variance threshold;
• and
• producing an updated background frame for pixels for which it is determined that corresponding pixel values of the differential matrix are less than the defined variance threshold, the producing being based at least in part on the converted retrieved video frame, the converted initial background frame, and a speed factor.

A system for producing a video abstract from video produced by a video camera, the system comprising a central processing unit (CPU) and a graphics processing unit (GPU), the CPU and GPU being configured to perform a method comprising:

• retrieving a frame of the video produced by the camera;
• updating a background frame based at least in part on the retrieved frame;
• performing a video division process to separate static frames from motive frames, the video division process comprising:
• retrieving, if it is determined that the retrieved frame is a static frame, a next frame of the video produced by the camera, updating the background frame, and repeating the video division process;
• and
• extracting, if it is determined that the retrieved frame is a motive frame, foreground data from the retrieved frame after conversion of the retrieved frame to a signed data type;
• determining whether an end of the video produced by the camera has been reached;
• retrieving, if it is determined that the end of the video produced by the camera has not been reached, a next frame of the video produced by the camera, updating the background frame, and repeating the video division process;
• merging, if it is determined that the end of the video produced by the camera has been reached, the foreground data with the updated background data after conversion of the foreground data and the updated background data to the signed data type;
• and
• generating a video abstract with the merged foreground and background data,
• wherein the updating of the background frame comprises:
• converting the retrieved video frame and an initial background frame to a signed data type;
• determining a differential matrix equal to an absolute value of pixel values of the retrieved video frame minus corresponding pixel values of the initial background frame;
• determining whether each pixel value of the differential matrix is less than a defined variance threshold;
• and
• producing an updated background frame for pixels for which it is determined that corresponding pixel values of the differential matrix are less than the defined variance threshold, the producing being based at least in part on the converted retrieved video frame, the converted initial background frame, and a speed factor.