Select Trends in Image , Video , and Multidimensional Signal Processing Gaurav

almost been synonymous with machine vision, which suggests processing of data from cameras operating in the visible range, the trend is to use multiple modalities of inputs that include sensors such as radar and microphone arrays. The trend is to tackle the problem using a layered model starting with the sensor outputs and working through other layers such as front-end processing, object localization, object recognition, context recognition, and spatiotemporal perception. Key technical challenges include: 1) developing robust real-world algorithms for midlevel tasks, 2) generating “complete” ontologies of scenes/scenarios of interest, and 3) discerning and describing events beyond the trained set. With respect to autonomous systems, there are further challenges such as computation, scalability over robot platforms, interfacing with ma chine intelligence, and human robot interface components. The field offers a rich variety of signal processing problems in areas such as multichannel processing involving multimodal sensor outputs, cue and behavior inference, and symbolic representations.