Background subtraction based on Local Shape
This is an incremental improvement for computer vision applications like video surveillance.
The paper tackles background subtraction by modeling local shape changes using self-similarity descriptors, resulting in complete foreground detection but with shadows often included.
We present a novel approach to background subtraction that is based on the local shape of small image regions. In our approach, an image region centered on a pixel is mod-eled using the local self-similarity descriptor. We aim at obtaining a reliable change detection based on local shape change in an image when foreground objects are moving. The method first builds a background model and compares the local self-similarities between the background model and the subsequent frames to distinguish background and foreground objects. Post-processing is then used to refine the boundaries of moving objects. Results show that this approach is promising as the foregrounds obtained are com-plete, although they often include shadows.