Search

Stereo matching is an inherently difficult problem due to ambiguous and noisy texture. The non-convexity and non-differentiability makes local linear (or quadratic) approximations poor, thereby preventing the use of standard local descent methods. Therefore recent methods are predominantly based on discretization and/or random sampling of some class of approximating surfaces (e.g. planes). While these methods are very efficient in generating a rough surface estimate, via either fusion of proposals or label propagation, the end result is usually not as smooth as desired. In this paper we show that, if the objective function is decomposed correctly, local refinement of candidate solutions can be performed using an ADMM approach. This allows searching over more general function classes, thereby resulting in visually more appealing smooth surface estimations.

This paper presents an effective feature representation method in the context of activity recognition. Efficient and effective feature representation plays a crucial role not only in activity recognition, but also in a wide range of applications such as motion analysis, tracking, 3D scene understanding etc. In the context of activity recognition, local features are increasingly popular for representing videos because of their simplicity and efficiency. While they achieve state-of-the-art performance with low computational requirements, their performance is still limited for real world applications due to a lack of contextual information and models not being tailored to specific activities. We propose a new activity representation framework to address the shortcomings of the popular, but simple bag-of-words approach. In our framework, first multiple instance SVM (mi-SVM) is used to identify positive features for each action category and the k-means algorithm is used to generate a codebook. Then locality-constrained linear coding is used to encode the features into the generated codebook, followed by spatio-temporal pyramid pooling to convey the spatio-temporal statistics. Finally, an SVM is used to classify the videos. Experiments carried out on two popular datasets with varying complexity demonstrate significant performance improvement over the base-line bag-of-feature method.

Corner detection has shown its great importance in many computer vision tasks. However, in real-world applications, noise in the image strongly affects the performance of corner detectors. Few corner detectors have been designed to be robust to heavy noise by now, partly because the noise could be reduced by a denoising procedure. In this paper, we present a corner detector that could find discriminative corners in images contaminated by noise of different levels, without any denoising procedure. Candidate corners (i.e., features) are firstly detected by a modified SUSAN approach, and then false corners in noise are rejected based on their local characteristics. Features in flat regions are removed based on their intensity centroid, and features on edge structures are removed using the Harris response. The detector is self-adaptive to noise since the image signal-to-noise ratio (SNR) is automatically estimated to choose an appropriate threshold for refining features. Experimental results show that our detector has better performance at locating discriminative corners in images with strong noise than other widely used corner or keypoint detectors.

The mean shift tracker has achieved great success in visual object tracking due to its efficiency being nonparametric. However, it is still difficult for the tracker to handle scale changes of the object. In this paper, we associate a scale adaptive approach with the mean shift tracker. Firstly, the target in the current frame is located by the mean shift tracker. Then, a feature point matching procedure is employed to get the matched pairs of the feature point between target regions in the current frame and the previous frame. We employ FAST-9 corner detector and HOG descriptor for the feature matching. Finally, with the acquired matched pairs of the feature point, the affine transformation between target regions in the two frames is solved to obtain the current scale of the target. Experimental results show that the proposed tracker gives satisfying results when the scale of the target changes, with a good performance of efficiency.

The location of previously unseen and unregistered individuals in complex camera networks from semantic descriptions is a time consuming and often inaccurate process carried out by human operators, or security staff on the ground. To promote the development and evaluation of automated semantic description based localisation systems, we present a new, publicly available, unconstrained 110 sequence database, collected from 6 stationary cameras. Each sequence contains detailed semantic information for a single search subject who appears in the clip (gender, age, height, build, hair and skin colour, clothing type, texture and colour), and between 21 and 290 frames for each clip are annotated with the target subject location (over 11,000 frames are annotated in total). A novel approach for localising a person given a semantic query is also proposed and demonstrated on this database. The proposed approach incorporates clothing colour and type (for clothing worn below the waist), as well as height and build to detect people. A method to assess the quality of candidate regions, as well as a symmetry driven approach to aid in modelling clothing on the lower half of the body, is proposed within this approach. An evaluation on the proposed dataset shows that a relative improvement in localisation accuracy of up to 21 is achieved over the baseline technique.

We present a novel technique for the recognition of complex human gestures for video annotation using accelerometers and the hidden Markov model. Our extension to the standard hidden Markov model allows us to consider gestures at different levels of abstraction through a hierarchy of hidden states. Accelerometers in the form of wrist bands are attached to humans performing intentional gestures, such as umpires in sports. Video annotation is then performed by populating the video with time stamps indicating significant events, where a particular gesture occurs. The novelty of the technique lies in the development of a probabilistic hierarchical framework for complex gesture recognition and the use of accelerometers to extract gestures and significant events for video annotation.

In order to enable high-level semantics-based video annotation and interpretation, we tackle the problem of automatic decomposition of motion pictures into meaningful story units, namely scenes. Since a scene is a complicated and subjective concept, we first propose guidelines from film production to determine when a scene change occurs in film. We examine different rules and conventions followed as part of Film Grammar to guide and shape our algorithmic solution for determining a scene boundary. Two different techniques are proposed as new solutions in this paper. Our experimental results on 10 full-length movies show that our technique based on shot sequence coherence performs well and reasonably better than the color edges-based approach.

Gait classification is a developing research area, particularly with regards to biometrics. It aims to use the distinctive spatial and temporal characteristics of human motion to classify differing activities. As a biometric, this extends to recognising different people by the heterogeneous aspects of their gait. This research aims to use a modified deformable model, the temporal PDM, to distinguish the movements of a walking and miming person. The movement of 2D points on the moving form is used to provide input into the model and classify the type of gait present.

The methods for visual learning that compute a space of eigenvectors by Principal Component Analysis (PCA) traditionally require a batch computation step. Since this leads to potential problems when dealing with large sets of images, several incremental methods for the computation of the eigenvectors have been introduced. However, such learning cannot be considered as an on-line process, since all the images are retained until the final step of computation of space of eigenvectors, when their coefficients in this subspace are computed. In this paper we propose a method that allows for simultaneous learning and recognition. We show that we can keep only the coefficients of the learned images and discard the actual images and still are able to build a model of appearance that is fast to compute and open-ended. We performed extensive experimental testing which showed that the recognition rate and reconstruction accuracy are comparable to those obtained by the batch method.