Background understanding in video data

摘要

Long-term understanding of background modeling includes determining first and second dimension gradient model derivatives of image brightness data of an image pixel along respective dimensions of two-dimensional, single channel image brightness data of a static image scene. The determined gradients are averaged with previous determined gradients of the image pixels, and with gradients of neighboring pixels as a function of their respective distances to the image pixel, the averaging generating averaged pixel gradient models for each of a plurality of pixels of the video image data of the static image scene that each have mean values and weight values. Background models for the static image scene are constructed as a function of the averaged pixel gradients and weights, wherein the background model pixels are represented by averaged pixel gradient models having similar orientation and magnitude and weights meeting a threshold weight requirement.

主权项

1. A method of providing a service for long-term understanding of background modeling, the method comprising providing:
a gradient determiner that uses features extracted from input video data to determine dimensional gradient models for pixel image data of the input video and define average image pixel gradient models for each pixel of the pixel image data by averaging the determined gradients with previous gradients of the each pixel, and also with gradients of neighboring pixels as a function of their distance to the pixel; a background modeler that constructs and updates background pixel models for a static image scene of the input video data by using the averaged pixel gradient data that have similar orientation and magnitude for each of a plurality of pixel model sets, wherein each pixel model set is associated with a weight that determines if the each pixel model set represents background or non-background data; and a foreground estimator that uses the constructed and updated background models to estimate foreground areas in the static image scene of the input video data by comparing the averaged pixel gradients with corresponding ones of the background models, wherein a pixel is determined to be a foreground and not a background pixel in response to a distance from the averaged pixel gradient data to a corresponding background model pixel set equaling or exceeding a threshold distance value, or to the distance being less than the threshold but closer to another foreground model pixel set than to the corresponding background model pixel set.