In recent years, research in human action recognition has advanced on multiple fronts to address various types of actions including simple, isolated actions in staged data (e.g., KTH dataset), complex actions (e.g., Hollywood dataset), and naturally occurring actions in surveillance videos (e.g, VIRAT dataset). Several techniques including those based on gradient, flow, and interest-points, hav...

Moment invariants are popular descriptors for real valued functions. Their independence from certain transformations makes them a powerful tool for the recognition of patterns and shapes. It has recently been demonstrated that the basic ideas can also be transferred to vector valued functions. Vector moment invariants can be used to define and search for interesting flow structures. A generaliz...

We present a method for unsupervised learning of classes of motions in video. We project optical flow fields to a complete, orthogonal, a-priori set of basis functions in a probabilistic fashion, which improves the estimation of the projections by incorporating uncertainties in the flows. We then cluster the projections using a mixture of feature-weighted Gaussians over optical flow fields. The...

We demonstrate the control of droplet sizes by an ac voltage applied across microelectrodes patterned around a flow-focusing junction. The electrodes do not come in contact with the fluids to avoid electrochemical effects. We found several regimes of droplet production in electric fields, controlled by the connection of the chip, the conductivity of the dispersed phase and the frequency of the ...

Due to the amount of data nowadays, automatic detection, classification and visualization of features is necessary for a thorough inspection of flow data sets. Pattern matching using vector valued templates has already been applied successfully for the detection of features. In this paper, the approach is extended to automatically compute feature based segmentations of flow data sets. Different...

Accurate optical flow estimation of motion fields is crucial for video super-resolution algorithms. Existing algorithms for optical flow calculation may produce erroneous motion fields on complex dynamic scenes with multiple moving, occluding non-rigid objects. In this paper we propose to use so called optical flow reliability weights in order to reduce impact of erroneous motion vector estimat...

valuable as a source of spatial information. Current techniques provide flow fields which are noisy and sparse, mak ing the recovery of spatial proper ties difficul t at best. This paper describes a technique for locating discontinuities in optical flow, which typically correspond to object boundaries. A simple blurring interpolator is used to smooth out noise and produce denser fields. Discont...

To identify the content of motion sequences such as gestures and video sequences, we propose a compact representation of flow fields using a basis of complex Zernike polynomials. Moments derived from projection of these polynomials have been used in recognition. Previously, local descriptions learned from examples, or specified a priori, have been used to identify local flow events. Affine expa...

We derive a formal link between temporally weighted frame differences, or disturbance fields, which carry limited information commonly used for motion detection, and the optic flow, which carries rich information on local image motion. We use this to formulate a novel, simple, near-recursive optic flow algorithm based on a recursive-filter formulation. Most quantities involved are computed recu...

Many natural phenomena involve dynamic fluids whose motion differs radically from that of rigid bodies. This paper presents a novel approach for estimating fluid motion fields. It first estimates a local flow probability distribution function at each pixel using the STAR model and the data from a spatio-temporal neighborhood. It then feeds the set of distribution functions into a global optimiz...