Some aspects of information processing in biological vision

An attempt is made to present some challenging problems (mainly to the technically minded researchers) in the development of computational models for certain (visual) processes which are executed with, apparently, deceptive ease by the human visual system. However, in the interest of simplicity (and with a nonmathematical audience in mind), the presentation is almost completely devoid of mathematical formalism. Some of the findings in biological vision are presented in order to provoke some approaches to their computational models, The development of ideas is not complete, and the vast literature on biological and computational vision cannot be reviewed here. A related but rather specific aspect of computational vision (namely, detection of edges) has been discussed by Zucker, who brings out some of the difficulties experienced in the classical approaches.Space limitations here preclude any detailed analysis of even the elementary aspects of information processing in biological vision, However, the main purpose of the present paper is to highlight some of the fascinating problems in the frontier area of modelling mathematically the human vision system.

Dynamics of 3D view invariance in monkey inferotemporal cortex

Rotations in depth are challenging for object vision because features can appear, disappear, be stretched or compressed. Yet we easily recognize objects across views. Are the underlying representations view invariant or dependent? This question has been intensely debated in human vision, but the neuronal representations remain poorly understood. Here, we show that for naturalistic objects, neurons in the monkey inferotemporal (IT) cortex undergo a dynamic transition in time, whereby they are initially sensitive to viewpoint and later encode view-invariant object identity. This transition depended on two aspects of object structure: it was strongest when objects foreshortened strongly across views and were similar to each other. View invariance in IT neurons was present even when objects were reduced to silhouettes, suggesting that it can arise through similarity between external contours of objects across views. Our results elucidate the viewpoint debate by showing that view invariance arises dynamically in IT neurons out of a representation that is initially view dependent.

Learning Feature Trajectories Using Gabor Filter Bank for Human Activity Segmentation and Recognition

We describe a novel method for human activity segmentation and interpretation in surveillance applications based on Gabor filter-bank features. A complex human activity is modeled as a sequence of elementary human actions like walking, running, jogging, boxing, hand-waving etc. Since human silhouette can be modeled by a set of rectangles, the elementary human actions can be modeled as a sequence of a set of rectangles with different orientations and scales. The activity segmentation is based on Gabor filter-bank features and normalized spectral clustering. The feature trajectories of an action category are learnt from training example videos using dynamic time warping. The combined segmentation and the recognition processes are very efficient as both the algorithms share the same framework and Gabor features computed for the former can be used for the later. We have also proposed a simple shadow detection technique to extract good silhouette which is necessary for good accuracy of an action recognition technique.

Visualization of in situ intracellular aggregation of two cataract-associated human gamma-crystallin mutants: Lose a tail, lose transparency

PURPOSE. To understand the molecular features underlying autosomal dominant congenital cataracts caused by the deletion mutations W156X in human gamma D-crystallin and W157X in human gamma C-crystallin. METHODS. Normal and mutant cDNAs (with the enhanced green fluorescent protein [EGFP] tag in the front) were cloned into the pEGFP-C1 vector, transfected into various cell lines, and observed under a confocal microscope for EGFP fluorescence. Normal and W156X gamma D cDNAs were also cloned into the pET21a(+) vector, and the recombinant proteins were overexpressed in the BL-21(DE3) pLysS strain of Escherichia coli, purified, and isolated. The conformational features, structural stability, and solubility in aqueous solution of the mutant protein were compared with those of the wild type using spectroscopic methods. Comparative molecular modeling was performed to provide additional structural information. RESULTS. Transfection of the EGFP-tagged mutant cDNAs into several cell lines led to the visualization of aggregates, whereas that of wild-type cDNAs did not. Turning to the properties of the expressed proteins, the mutant molecules show remarkable reduction in solubility. They also seem to have a greater degree of surface hydrophobicity than the wild-type molecules, most likely accounting for self-aggregation. Molecular modeling studies support these features. CONCLUSIONS. The deletion of C-terminal 18 residues of human gamma C-and gamma D-crystallins exposes the side chains of several hydrophobic residues in the sequence to the solvent, causing the molecule to self-aggregate. This feature appears to be reflected in situ on the introduction of the mutants in human lens epithelial cells.

A vision modeling framework for DHM using geometrically estimated FoV

Digital human modeling (DHM) involves modeling of structure, form and functional capabilities of human users for ergonomics simulation. This paper presents application of geometric procedures for investigating the characteristics of human visual capabilities which are particularly important in the context mentioned above. Using the cone of unrestricted directions through the pupil on a tessellated head model as the geometric interpretation of the clinical field-of-view (FoV), the results obtained are experimentally validated. Estimating the pupil movement for a given gaze direction using Listing's Law, FoVs are re-computed. Significant variation of the FoV is observed with the variation in gaze direction. A novel cube-grid representation, which integrated the unit-cube representation of directions and the enhanced slice representation has been introduced for fast and exact point classification for point visibility analysis for a given FoV. Computation of containment frequency of every grid-cell for a given set of FoVs enabled determination of percentile-based FoV contours for estimating the visual performance of a given population. This is a new concept which makes visibility analysis more meaningful from ergonomics point-of-view. The algorithms are fast enough to support interactive analysis of reasonably complex scenes on a typical desktop computer. (C) 2011 Elsevier Ltd. All rights reserved.

Human Action Recognition in Compressed Domain using PBL-McRBFN Approach

Large variations in human actions lead to major challenges in computer vision research. Several algorithms are designed to solve the challenges. Algorithms that stand apart, help in solving the challenge in addition to performing faster and efficient manner. In this paper, we propose a human cognition inspired projection based learning for person-independent human action recognition in the H.264/AVC compressed domain and demonstrate a PBL-McRBEN based approach to help take the machine learning algorithms to the next level. Here, we use gradient image based feature extraction process where the motion vectors and quantization parameters are extracted and these are studied temporally to form several Group of Pictures (GoP). The GoP is then considered individually for two different bench mark data sets and the results are classified using person independent human action recognition. The functional relationship is studied using Projection Based Learning algorithm of the Meta-cognitive Radial Basis Function Network (PBL-McRBFN) which has a cognitive and meta-cognitive component. The cognitive component is a radial basis function network while the Meta-Cognitive Component(MCC) employs self regulation. The McC emulates human cognition like learning to achieve better performance. Performance of the proposed approach can handle sparse information in compressed video domain and provides more accuracy than other pixel domain counterparts. Performance of the feature extraction process achieved more than 90% accuracy using the PTIL-McRBFN which catalyzes the speed of the proposed high speed action recognition algorithm. We have conducted twenty random trials to find the performance in GoP. The results are also compared with other well known classifiers in machine learning literature.

Assessment of driver vision functions in relation to their crash involvement in India

Among the human factors that influence safe driving, visual skills of the driver can be considered fundamental. This study mainly focuses on investigating the effect of visual functions of drivers in India on their road crash involvement. Experiments were conducted to assess vision functions of Indian licensed drivers belonging to various organizations, age groups and driving experience. The test results were further related to the crash involvement histories of drivers through statistical tools. A generalized linear model was developed to ascertain the influence of these traits on propensity of crash involvement. Among the sampled drivers, colour vision, vertical field of vision, depth perception, contrast sensitivity, acuity and phoria were found to influence their crash involvement rates. In India, there are no efficient standards and testing methods to assess the visual capabilities of drivers during their licensing process and this study highlights the need for the same.