head image

Research - Image Processing

The core of our approach is application of methods that decrease image information in one frame without losing information that is necessary to compute required image features.

For example, for foreground detection it is not always necessary to work in the original full image resolution as the foregroung objects of interest usually have sizes bigger than individual pixels. The key requirement then is to preserve information contributed by individual pixels while down-scaling the original image and computing the foreground mask in the reduced resolution.