S. M. Mahbubur Rahman

Detection and Classification of Vehicles From Video Using Multiple Time-Spatial Images

Detection and classification of vehicles are two of the most challenging tasks of a video-based intelligent transportation system. Traditional detection and classification methods are computationally highly expensive and become unsuccessful in many cases such as occlusion among the vehicles and when differences between pixel intensities of vehicles and backgrounds are small. In this paper, a novel detection and classification method is proposed using multiple time-spatial images (TSIs), each obtained from a virtual detection line on the frames of a video. Such a use of multiple TSIs provides the opportunity to identify the latent occlusions among the vehicles and to reduce the dependencies of the pixel intensities between the still and moving objects to increase the accuracy of detection performance as well as to achieve an improved classification performance. In order to identify the class of a particular vehicle, a two-step k nearest neighborhood classification scheme is proposed by utilizing the shape-based, shape-invariant, and texture-based features of the segmented regions corresponding to the vehicle appeared in appropriate frames that are determined from the TSIs of the video. Extensive experimentations are carried out in vehicular traffics of varying environments to evaluate the detection and classification performance of the proposed method, as compared with the existing methods. Experimental results demonstrate that the proposed method provides a significant improvement in counting and classifying the vehicles in terms of accuracy and robustness alongside a substantial reduction of execution time, as compared with that of the other methods.

On the selection of 2D Krawtchouk moments for face recognition

Sparse representation of images using orthogonal two-dimensional Krawtchouk moments (2D KCMs) for face recognition is motivated by their ability to capture region-based higher-order hidden nonlinear structures from discrete coordinates of finitely supported images and the invariance of affine transformations of these moments to common geometric distortions. This paper presents the effectiveness of selecting the discriminatory set of KCMs as the global and local face features as opposed to traditional features obtained from heuristic choice of fixed-order moments or projection of the moments for recognizing an identity. The selection of significantly sparse 2D KCM-based features according to the proposed approach results in highly efficient face recognition method as compared to the other methods that use orthogonal moments such as the 2D Zernike, 2D Tchebichef or 2D Gaussian-Hermite. Experiments on challenging databases (viz., FRGC and CK-AUC) and comparisons with the well established projection, texture, and moment-based methods indicate superior recognition performance in terms of mean accuracy and robustness of the proposed holistic- or hybrid-type discriminative KCM-based method, especially when sample sizes are small and the intraclass faces have significant variations due to expressions. HighlightsHybrid-type face recognition using discriminative selection of Krawtchouk moments.A comparative study of different moment-based discriminative face features.Experiments on two challenging databases show superiority of the proposed selection.Fisher-scoring is preferable to LDA projection for moment-based face recognition.Effective facial parts are identified for recognition of expression variant faces.

Differential components of discriminative 2D Gaussian–Hermite moments for recognition of facial expressions

Abstract This paper deals with a new expression recognition method by representing facial images in terms of higher-order two-dimensional orthogonal Gaussian–Hermite moments (GHMs) and their geometric invariants. Only the moments having high discrimination power are selected as a set of features for expressions. To obtain the differentially expressive components of the moments, the discriminative GHMs are projected on to a new expression-invariant subspace using the correlations among the neutral faces. Features obtained from the discriminative moments and differentially expressive components of the moments are used to recognize an expression using the well-known support vector machine classifier. Experimental results presented are obtained from commonly-referred databases such as the CK-AUC, FRGC, and MMI that have posed or spontaneous expressions as well as the GENKI database that has expressions in-the-wild. Experiments on mutually exclusive subjects reveal that the performance of expression recognition of the proposed method is significantly better than that of the existing or similar methods, which use the local or patch-based high dimensional binary patterns, directional number patterns generated from derivatives of Gaussian, Gabor- or other moment-based features.

Image fusion technique using multivariate statistical model for wavelet coefficients

Wavelet-based image fusion techniques have been highly successful in combining important features such as edges and textures of source images. In this work, a new discrete wavelet transform (DWT)-based fusion algorithm is proposed using a locally-adaptive multivariate statistical model for the wavelet coefficients of the source images as well as that of the fused image. The multivariate model is proposed based on the fact that the DWT coefficients of source images are correlated not only with each other but also with the fused image. By using this model as a joint prior function, an estimate of the fused coefficients is derived via the Bayesian maximum a posteriori estimation technique. Experimental results show that performance of the proposed fusion method is better than that of the other methods in terms of commonly-used metrics such as structural similarity, peak signal-to-noise ratio, and cross-entropy.

Detection and classification of vehicles from a video using time-spatial image

Detection and classification of vehicles are the most challenging tasks of a video-based intelligent transportation system. Traditional detection and classification methods are based on subtraction of estimated still backgrounds from a video to find out the moving objects. In general, these methods are computationally highly expensive, and in many cases show poor detection and classification performance, especially when differences between pixel intensities of vehicles and backgrounds are small. In this paper, we present a novel detection and classification method that employs an analysis of time-spatial image (TSI) obtained from a virtual line on the frames of a video so that the dependencies of pixel intensities of still and moving objects of the video may be reduced. First, the TSI is segmented to count the number of vehicles those cross the virtual line. Then, a feature-based classification scheme is proposed to classify these vehicles. The classification scheme utilizes the shape of the segmented regions of the TSI as well as that of appropriate frames of a video to extract the ceratin features of the moving objects. Experimental results on a number of real video sequences demonstrate that the proposed method provides higher accuracy in counting and classifying the vehicles as compared to that of the conventional background subtraction-based methods.

Low-complexity iris recognition method using 2D Gauss-Hermite moments

The authenticity and reliability of iris recognition-based biometric identification system is well-proven. Traditional iris recognition methods use expensive feature extraction algorithms and complex-valued IrisCodes that may hinder the development of a fast identification technique for multimodal biometric system. In this paper, a new set of computationally efficient real-valued features is proposed for recognition of iris patterns using the two dimensional higher-order Gauss-Hermite moments. The IrisCodes generated from the zero-crossings of these moment-based features are capable of capturing hidden nonlinear structures and are potentially invariant to distortions of iris patterns. Experimental results conducted on a generic data set consisting of iris images obtained from two well-known databases show that the proposed method provides encouraging performance. In particular, an acceptable recognition performance in terms of probability of detection for a given false alarm rate may be achieved by the proposed method with a significantly low-level of computational complexity.

Static hand gesture recognition using discriminative 2D Zernike moments

Hand gesture recognition plays a vital role in developing vision-based communication for human-computer interaction. This paper presents a novel static hand gesture recognition method using the two dimensional Zernike moments (2D ZMs) those are considered as effective features when patterns in images possess distortions due to rotation, scaling or viewing angle. The key contribution of this paper lies in the fact that a discriminative set of ZMs are used to represent features of the hand postures as opposed to traditional features obtained from heuristic choice of fixed-order moments. The orthogonal nature of the 2D ZMs allows the estimation of the discrimination power of the individual moments by using the inter- and intra-class variances of the features. The nearest neighbor classifier is employed on the discriminative ZMs (DZMs) to recognize the hand postures in a computationally efficient way. Experimental results on commonly-referred database show that the proposed DZM-based method provides recognition accuracies better than that provided by the conventional principal component analysis, Fourier descriptor or existing ZM-based methods.

Tracking-based detection of driving distraction from vehicular interior video

Distraction during driving is a growing concern for global road safety. Different activities impertinent to driving hinder the concentration of driver on road and often cause substantial damage to life and property. For making driving safe, an algorithm is proposed in this paper that is capable of detecting distraction during driving. The proposed algorithm tracks key body parts of the driver in video captured by a front camera. Euclidean distances between the tracking trajectories of body parts are used as representative features that characterize the state of distraction or attention of a driver. The well-known K-nearest neighbor classifier is applied for detecting distraction from the features extracted from body parts. The proposed method is compared with existing methods implementing tracking-based human action identification to corroborate its improved performance.

Low-Complexity Nonrigid Image Registration Using Feature-Based Diffeomorphic Log-Demons

Traditional hybrid-type nonrigid registration algorithm uses affine transformation or class-specific distortion parameters for global matching assuming linear-type deformations in images. In order to consider generalized and nonlinear-type deformations, this paper presents an approach of feature-based global matching algorithm prior to certain local matching. In particular, the control points in images are identified globally by the well-known robust features such as the SIFT, SURF, or ASIFT and interpolation is carried out by a low-complexity orthogonal polynomial transformation. The local matching is performed using the diffeomorphic Demons, which is a well-established intensity-based registration method. Experiments are carried out on synthetic distortions such as spherical, barrel, and pin-cushion in commonly referred images as well as on real-life distortions in medical images. Results reveal that proposed introduction of feature-based global matching significantly improves registration performance in terms of residual errors, computational complexity, and visual quality as compared to the existing methods including the log-Demons itself.

Gaussian-Hermite moment-based depth estimation from single still image for stereo vision

Abstract Depth information of objects plays a significant role in image-based rendering. Traditional depth estimation techniques use different visual cues including the disparity, motion, geometry, and defocus of objects. This paper presents a novel approach of focus cue-based depth estimation for still images using the Gaussian-Hermite moments (GHMs) of local neighboring pixels. The GHMs are chosen due to their superior reconstruction ability and invariance properties to intensity and geometric distortions of objects as compared to other moments. Since depths of local neighboring pixels are significantly correlated, the Laplacian matting is employed to obtain final depth map from the moment-based focus map. Experiments are conducted on images of indoor and outdoor scenes having objects with varying natures of resolution, edge, occlusion, and blur contents. Experimental results reveal that the depth estimated from GHMs can provide anaglyph images with stereo quality better than that provided by existing methods using traditional visual cues.