The problem we consider in this paper is to take a single two-dimensional image containing a human figure, locate the joint positions, and use these to estimate the body configuration and pose in three-dimensional space. The basic approach is to store a number of exemplar 2D views of the human body in a variety of different configurations and viewpoints with respect to the camera. On each of these stored views, the locations of the body joints (left elbow, right knee, etc.) are manually marked and labeled for future use. The input image is then matched to each stored view, using the technique of shape context matching in conjunction with a kinematic chain-based deformation model. Assuming that there is a stored view sufficiently similar in configuration and pose, the correspondence process will succeed. The locations of the body joints are then transferred from the exemplar view to the test shape. Given the 2D joint locations, the 3D body configuration and pose are then estimated using an existing algorithm. We can apply this technique to video by treating each frame independently—tracking just becomes repeated recognition. We present results on a variety of data sets.

We propose a new unsupervised learning method to obtain a layered pictorial structure (LPS) representation of an articulated object from video sequences. It will be seen that this is related in turn to methods for learning sprite based representations of an image. The method we describe involves a new generative model for performing segmentation on a set of images. Included in this model are the effects of motion blur and occlusion. An initial estimate of the parameters of the model is obtained by dividing the scene into rigidly moving components. The estimate of the matte of each part is refined using a variation of the α-expansion graph cut algorithm. This method has the advantage of achieving a strong local minimum over labels. Results are demonstrated on animals for which an articulated LPS representation is naturally suited. 1.

This paper presents an algorithm that categorises animal locomotive behaviour by combining detection and tracking of animal faces in wildlife videos. As an example, the algorithm is applied to lion faces.

Abstract not found

In this paper, we describe a system that can build 3D animal models and synthesize animations in 3D virtual environments. The model is constructed by 2D images captured by specific views. The animation is synthesised by using physical motion models of the animal and tracking data from image sequences. Finally, the user selects some points of the 3D world and a smooth and safe motion path, which passes by these points, is created. The main assumption of the 3D modelling is that the animal could be divided into parts whose normal sections are ellipses. Joints and angles between skeleton points are used in order to decrease models complexity. Using the above methodology, a snake, a lizard and a goat are reconstructed. 1.

Abstract—We present a probabilistic method for segmenting instances of a particular object category within an image. Our approach overcomes the deficiencies of previous segmentation techniques based on traditional grid conditional random fields (CRF), namely that 1) they require the user to provide seed pixels for the foreground and the background and 2) they provide a poor prior for specific shapes due to the small neighborhood size of grid CRF. Specifically, we automatically obtain the pose of the object in a given image instead of relying on manual interaction. Furthermore, we employ a probabilistic model which includes shape potentials for the object to incorporate top-down information that is global across the image, in addition to the grid clique potentials which provide the bottom-up information used in previous approaches. The shape potentials are provided by the pose of the object obtained using an object category model. We represent articulated object categories using a novel layered pictorial structures model. Nonarticulated object categories are modeled using a set of exemplars. These object category models have the advantage that they can handle large intraclass shape, appearance, and spatial variation. We develop an efficient method, OBJCUT, to obtain segmentations using our probabilistic framework. Novel aspects of this method include: 1) efficient algorithms for sampling the object category models of our choice and 2) the observation that a sampling-based approximation of the expected log-likelihood of the model can be increased by a single graph cut. Results are presented on several articulated (e.g., animals) and nonarticulated (e.g., fruits) object categories. We provide a favorable comparison of our method with the state of the art in object category specific image segmentation, specifically the methods of Leibe and Schiele and Schoenemann and Cremers.

Abstract not found

Films like Shrek, Madagascar, The Chronicles of Narnia and Charlotte’s web all have something in common: realistic quadruped animations. While the animation of animals has been popular for a long time, the technical challenges associated with creating highly realistic, computer generated creatures have been receiving increasing attention recently. The entertainment, education and medical industries have increased the demand for simulation of realistic animals in the computer graphics area. In order to achieve this, several challenges need to be overcome: gathering and processing data that embodies the natural motion of an animal – which is made more difficult by the fact that most animals cannot be easily motion-captured; build accurate kinematic models for animals, in particular with adapted animation skeletons; and develop either kinematic or physically-based animation methods, either embedding some a priori knowledge about the way that quadrupeds locomote and/or building on some example of real motion. In this state of the art report, we present an overview of the common techniques used to date for realistic quadruped animation. This includes an outline of the various ways that realistic quadruped motion can be achieved, through video-based acquisition, physics based models, inverse kinematics, or some combination of the above. The research presented represents a cross fertilisation of vision, graphics and interaction methods. Categories and Subject Descriptors (according to ACM CCS): Quadruped Animation, Behavioural Simulation 1.

This paper presents an algorithm for detection and tracking of animal faces in wildlife videos. As an example the algorithm is applied to lion faces. The detection algorithm is based on a human face detection method, utilising Haar-like features and AdaBoost classifiers. The face tracking is implemented using the Kanade-Lucas-Tomasi tracker and by applying a specific interest model to the detected face. By combining the two methods in a specific tracking model, a reliable and temporally coherent detection/tracking of animal faces is achieved. In addition to the detection of particular animal species, the information generated by the tracker can be used to boost the priors in the probabilistic semantic classification of wildlife videos

Abstract—During the past several years, face recognition in video has received significant attention. Not only the wide range of commercial and law enforcement applications, but also the availability of feasible technologies after several decades of research contributes to the trend. Although current face recognition systems have reached a certain level of maturity, their development is still limited by the conditions brought about by many real applications. For example, recognition images of video sequence acquired in an open environment with changes in illumination and/or pose and/or facial occlusion and/or low resolution of acquired image remains a largely unsolved problem. In other words, current algorithms are yet to be developed. This paper provides an up-to-date survey of video-based face recognition research. To present a comprehensive survey, we categorize existing video based recognition approaches and present detailed descriptions of representative methods within each category. In addition, relevant topics such as real time detection, real time tracking for video, issues such as illumination, pose, 3D and low resolution are covered. Keywords—Face recognition, video-based, survey I.