This paper presents a new model of human attention that allows salient areas to be extracted from video frames. As automatic understanding of video semantic content is still far fr...
We rephrase the problem of 3D reconstruction from images in terms of intersections of projections of orbits of custom built Lie groups actions. We then use an algorithmic method b...
A novel motion activity descriptor and its extraction from a compressed MPEG (MPEG-1/2) video stream are presented. The descriptor consists of two parts, a temporal descriptor and ...
The problemof structurefrom motion (SFM)is to extract the three-dimensionalmodel of a moving scene from a sequence of images. Though two images are sufficient to produce a 3D reco...
Abstract. We show how to exploit temporal periodicity of moving objects to perform 3D reconstruction. The collection of period-separated frames serve as a surrogate for multiple ri...