No abstract

No abstract

Work on real-time hand-gesture recognition for SAVI (Stereo Active Vision Interface) is presented. Based on the detection of frontal faces, image regions near the face are searched for the existence of skin-tone blobs. Each blob is evaluated to determine i f it is a hand held in a standard pose. A verification algorithm based on the responses oj elongated oriented3filter-s is used to decide whether a hand is present or not.Once a hand is detected, gestures are given by varying the number ofjingers visible. The hand is segmented using an algorithm which detects connected skin-tone blobs in the region of interest, and a medial axis transform (skeletonization) is applied. Analysis of the resulting skeleton allows detection of the number of3ngers visible, thus determining the gesture. The skeletonization is sensitive to strong shadows which may alter the detected morphology of the hand. Experimental results are given indicating good performance of the algorithm.

A Stereo Active Vision Interface is introduced which detects frontal faces in real world environments and performs particular active control tasks dependent on changes in the visual field. Firstly, connected skin colour regions in the visual scene are detected by applying a radial scanline algorithm. Secondly, facial features are searched for in the most salient skin colour region while the blob is tracked by the camera system. The facial features are evaluated and, based on the obtained results and the current state of the system, particular actions are performed. The SAVI system is thought of as a smart user interface for teleconferencing, telemedicine, and distance learning. The system is designed as a Perception-Action-Cycle (PAC), processing sensory data of different kinds and qualities. Both the vision module and the head motion control module work at frame rate. Hence, the system is able to react instantaneously to changing conditions in the visual scene.

We propose a denoising scheme to restore images degraded by CCD noise. Typically, restoration algorithms assume a linear mapping between the incident light space and image space. However, in practice a camera response function performs a non-linear mapping on the sensor output and as a result the sensor noise model becomes more complex in the image space. In this paper, we correct for non-linearity by mapping the corrupted image into "light space", where the relationship between the incident light and light space values is linear. To reduce the sensor noise we accurately model the CCD sensor noise by using the Photon Transfer Curve. We then develop a combination of adaptive filters based on the estimated noise model in light space. Our adaptive system demonstrates efficient noise removal performance in uniform regions, while preserving edges and fine details.