On the basis of feature points pairing, a scale-invariant feature matching method is proposed in this study. The distance between two features is used to compute feature pairs' support region size, which is different from the methods using detectors to provide information to find the support region. Moreover, to achieve rotation invariance, a sub-region division method based on intensity order is introduced. For comparison to the popular descriptors scale-invariant feature transform and speeded-up robust features, the authors also choose the detected points by difference of Gaussian and fast Hessain detectors as feature points to start the authors' method. Additional experiments compare the reported method with similar proposed methods, such as Tell's and Fan's. The experimental results show that the authors' proposed descriptor outperforms the popular descriptors under various image transformations, especially on images with scale and viewpoint transformations.

Traditional studies in vision-based hand gesture recognition remain rooted in view-dependent representations, and hence users are forced to be fronto-parallel to the camera. To solve this problem, view-invariant gesture recognition aims to make the recognition result independent of viewpoint changes. However, in current works the view-invariance is achieved at the price of mixing different gesture patterns that have similar trajectory curve shape but different semantic meanings. For example, the gesture ‘push’ can be mistaken as ‘drag’ from another viewpoint. To address this shortcoming, in this study, the authors use a shape descriptor to extract the view-invariant features of a three-dimensional (3D) trajectory. As the shape features are invariant to omnidirectional viewpoint changes, the orientation features are then added into weight different rotation angles so that similar trajectory shapes are better separated. The proposed method was conducted on two different databases, including a popular Australian Sign Language database and a challenging Kinect Hand Trajectory database. Experimental results show that the proposed algorithm achieves a higher average recognition rate than the state-of-the-art approaches, and can better distinguish confusing gestures while meeting the view-invariant condition.

Nowadays, the number of digital images has increased so that the management of this volume of data needs an efficient system for browsing, categorising and searching. Automatic image annotation is designed for assigning tags to images for more accurate retrieval. Non-negative matrix factorisation (NMF) is a traditional machine learning technique for decomposing a matrix into a set of basis and coefficients under the non-negative constraints. In this study, the authors propose a two-step algorithm for designing an automatic image annotation system that employs the NMF framework for its first step and a variant of K-nearest neighbourhood as its second step. In the first step, a new multimodal NMF algorithm is proposed to extract the latent factors which reflect the content of images. This is done by jointly factorising the visual and textual data feature matrices so that they have close representation, although not necessarily the same. In the second step, after mapping images to the latent factors space a few tags are predicted for the new images based on a weighted average of similar data. They evaluated the performance of the proposed method and compared it to the state-of-the-art literature. Comparison results demonstrate the effectiveness and potential of the proposed method in image annotation applications.

Region covariance descriptor that fuses multiple features compactly has proven to be very effective for visual tracking. While working effectively, the exhaustive global search strategy of covariance tracking is still inefficient, and there is much room for improvement. It may cause inconsecutive tracking trajectory and distraction. A suitable region similarity metric for covariance matching between the candidate object region and a given appearance template is of much importance. However, the computational burden of the metric, especially for large matrices under Riemannian space, may hinder its application in gradient-based algorithms. In this study, the authors propose an algorithm which, by minimising the metric function, exploits an efficient conjugate gradient method to iteratively search the best matched candidate, and determines the search step size by non-monotonic liner strategy. Then, an inferential reasoning in view of new efficient metric is derived for the gradient-based algorithm. The authors test the proposed tracking method on test baseline dataset. Both quantitative and qualitative results demonstrate the effectiveness of the proposed algorithm compared with other state-of-the-art methods.

Occlusion is one of the most challenging problems in many video processing applications such as surveillance, gait recognition, activity recognition and so on. Attempts have been made to develop algorithms for handling occlusion and evaluate their performance on various datasets. However, these studies are subjective in nature and the datasets are hardly characterised in terms of the level of occlusion, thereby precluding any form of quantitative comparison of performance. This shows a compelling need to design an explicit, unambiguous and quantitative model, which should be able to objectively represent occlusion in a video. This study proposes an occlusion model based on the position and pose uncertainties of the moving subjects in a video. The proposed occlusion model is able to characterise the level of occlusion present in a video. It is also employed to synthetically generate occlusion for walking sequences, thus providing a direction for controlled dataset generation against which human identification algorithms can be tested. Given an input video with a subject moving without any occlusion, a particle swarm optimisation-based parameter estimation methodology is presented that generates the desired level of occlusion. The proposed approaches have been tested on the TUM-IITKGP and PETS2010 datasets. Finally, as an application, the occlusion model has been used to generate an occluded gait datasets and the performances of different gait recognition algorithms have been compared under varying levels of occlusion.

Although multi-objects tracking has been improved significantly, tracking multiple aircrafts with nearly the same appearance remains a difficult task, especially when a significant pose changes and long-time occlusions occur in the complex environment. In this study, the authors propose a new multi-aircrafts tracker based on a structured support vector machine (SVM) and an intra-frame scale-invariant feature transform feature matching. The structured SVM-based model adapts to the appearance change well, but confuses different aircrafts when occlusions between aircrafts occur. To handle occlusions, an intra-frame matching method is applied to separate different aircrafts by matching points into different clusters. Moreover, to remove the mismatching caused by the cluttered background, the spatial–temporal constraint is applied to help improve the performance of the intra-frame feature matching. As there is no dataset to evaluate a multi-aircrafts tracker, they select eighteen challenging videos and manually annotate the ground truth, forming the first multi-aircrafts tracking dataset. The experiments in the dataset demonstrate that the author's tracker outperforms the state-of-the-art trackers in multi-aircrafts tracking.

The authors consider the general problem of graph clustering. Graph clustering manipulates the graph-based data structure and the entries of the solution vectors are only allowed to take non-negative discrete values. Finding the optimal solution is NP-hard, so relaxations are usually considered. Spectral clustering retains the orthogonality rigorously but ignores the non-negativity and discreteness of the solution. Sym non-negative matrix factorisation can retain the non-negativity rigorously but it is hard to reach the orthogonality. In this study, they proposed a novel method named congruent approximate graph clustering (CAC), which can retain the non-negativity rigorously and can reach the orthogonality properly by congruency approximation. Furthermore, the solution obtained by CAC is sparse, which is approximate with the ideal discrete solution. Experimental results on several real image benchmark datasets indicate that CAC achieves encouraging results compared with state-of-the-art methods.

In this study, the authors present an integrated approach for recovering both the intrinsic and extrinsic parameters of a camera from the silhouettes and feature point tracking of an object in a turntable image sequence. Their motivation in taking the integrated approach is to solve the problem of obtaining the Euclidean three-dimensional (3D) reconstruction of cereal roots growing in gellan gum or objects immersed in water and imaged under turntable motion. The problem of self-calibration by previous approaches is especially difficult in this case as the initialisation of the rotation axis from the bi-tangent lines to the surface of revolution is not feasible. The conics projected from the circular trajectories of root tips are highly eccentric. Their approach is to initialise the axis of rotation l _s from the centre of the conics fitted to feature point trajectory. An estimate of l _s is then iteratively estimated by minimising an error function in estimating the harmonic homology introduced by the surface of symmetry. They compare the results of their approach to those of the maximum-likelihood estimation-based approach to conic fitting of point feature trajectories. They show results of real 3D reconstruction of roots, which are detailed enough for phenotypic analysis and are better both quantitatively and qualitatively than those using just feature point tracking.

Robust object tracking is a challenging task because of factors such as pose variation, illumination changes, abrupt motion and background clutter across the video sequence. With the introduction of the compressive sensing theory, researchers are provided with a new and effective way of real-time object tracking. In this study, an enhanced fast compressive tracking based on an adaptive measurement matrix is presented, which the authors have named ‘adaptive fast compressive tracking’ (AFCT). The sparsity of the matrix and the number of columns are adaptively determined according to the dimension of the Haar-like feature. This measurement matrix is fixed once it has been calculated when selecting a tracked rectangle region in the first frame. Unlike most of the existing compressive trackers, the proposed method adopts a different adaptive measurement matrix for a different targeting object. Compared with the fast compressive tracking (FCT), each measurement element contains more information for the original signal. As a result, stable object tracking is achieved by using fewer measurement elements. The proposed AFCT method can run in real time and outperforms FCT on many challenging video sequences in terms of efficiency, accuracy and robustness.

This study presents a new Perona–Malik (PM) model which is based on new non-local means (NLM) theory for an image denoising. In the proposed model, the intensity, space position and gradient information are introduced into NLM model, which weakens the staircasing effect and preserves edge in a processed image. The new PM model enhances the denoising capability, and decreases fine characteristics to be over-smoothed. Comparative experiments show that the denoising capability of the proposed model is superior to the other three existing models.

Local binary patterns (LBPs) are a well-known operator that shows the ability for rotation and scale invariant texture classification. A recent extension of this operator is the pyramid transform domain approach on LBPs (PLBP). Obtaining more accuracy by using more pyramid representations is an important result of PLBP, which increases not only feature dimensionality, but also classification computational time (CT). This study illustrates that more pyramid image representations will not improve the performance of the PLBP. We evaluate efficient levels of representation for the PLBP descriptor. In addition, the authors propose some feature selection approaches, such as the multi-level and multi-resolution (ML + MR) approach and the ML, MR and multi-band (ML + MR + MB) approach and discuss their efficiency and CT. Experimental results show that the proposed feature selection approaches improve the accuracy of texture classification with fewer pyramid image representations. In addition, replacing the Chi-2 similarity measurement with Czekannowski improves the accuracy of texture classification.

Both Shannon and Kolmogorov–Chaitin–Solomonoff (KCS) information models fail to measure meaningful information in images. Pictures of a cow and correlated noise can both have the same Shannon and KCS information, but only the image of the cow has meaning. The application of ‘algorithmic specified complexity’ (ASC) to the problem of distinguishing random images, simple images and content-filled images is explored. ASC is a model for measuring meaning using conditional KCS complexity. The ASC of various images given a context of a library of related images is calculated. The ‘portable network graphic' (PNG) file format’s compression is used to account for typical redundancies found in images. Images which containing content can thereby be distinguished from those containing simply redundancies, meaningless or random noise.

Three-dimensional (3D) motion analysis of dynamic body organs using volumetric images is of increasing interest in different computer vision applications. A number of methods for estimation of 3D optical flow in those images have been developed in recent years. However, theoretical limits of 3D optical flow-based motion estimation and segmentation are yet to be analysed. In this study, a statistical analysis of 3D optical flow is presented and the results are used to predict the separability of local 3D motions. Experimental results, using both synthetic and real images, demonstrate the applicability of the proposed analysis to predict the separability of two motions in terms of the parameters quantifying their relative motion and the scale of measurement noise.

Image dehazing is an important issue that interests both image processing and computer vision. In this study, image dehazing is modelled as an example-based learning problem, and a novel dehazing algorithm using two-dimensional (2D) canonical correlation analysis (CCA) is proposed. By assuming that the hazy-free image patches are smooth and the pixel intensities in the same patch are approximate to constant, the authors deduce an underlying linear correlation between the observed hazy image patches and corresponding transmission patches. By maximising the correlation between the patch-pairs of hazy image and corresponding transmission map, 2D CCA is able to learn a subspace to reconstruct the reliable transmission. Thus, given a test hazy image, the transmission map is aggregated by the nearest neighbour patches in the subspace and then globally refined by a local mean adaptive guided filter. The final hazy-free image is obtained by using the dichromatic atmospheric model. Experimental results demonstrate the efficiency of the proposed method in single image dehazing.

A robust ellipse detector is proposed. The detector preprocesses the edge map by removing all the isolated points and conjunction points, and exploits polygonal curve to extract the elliptical arcs. A non-iterative geometric distance computation method is presented to serve a criterion which identifies the elliptical arcs belonging to the same ellipse by likelihood ratio test and fit ellipses to those merged arcs. The authors test their algorithm on both synthetic and real images, and the experimental results show a good performance of their algorithm.

This study presents a new local feature matching approach that relies upon Reeb graph (RG)-based representation as well as a simple and accurate similarity estimation. The central contribution of this work is to reinforce the topological consistency conditions of the graph-based description. Formally, the RGs are enriched with geometry signatures based on parameterisation approaches. After RG construction, the shape is segmented into Reeb charts of controlled topology mapped to its canonical planar domain. Then, two stretching signatures, corresponding to the area and angle distortion, are determined and taken as three-dimensional-shape descriptor. The similarity estimation is performed in two steps. The first one consists in forming the pairs of similar Reeb charts, according to the minimal distance between their corresponding signatures. The second step is to measure the global similarity which quantifies the similitude degree between all the matched Reeb charts. Retrieval experiments conducted on four publicly available databases have shown that the proposed matching scheme yields satisfactory results. Among observations, it can be noticed that despite its rapidity, the method provides an overall retrieval efficiency gain compared to very recent state-of-the-art methods.

To solve the problem of low efficiency and colour distortion in several typical tone mapping operators for high dynamic range (HDR) images, the authors propose a novel tone mapping algorithm based on fast image decomposition and multi-layer fusion. An input HDR image is firstly decomposed into a base layer and a detail layer by an improved guided filtering method. For the base layer, its dynamic range is compressed by the simulated camera response function. For the detail layer, it is enhanced to produce more fine structures and reduce halo effect by applying the guided image filter. The colour balance correction method is adopted to suppress colour distortion. The experiments on HDR images demonstrate that the proposed technique could bring better brightness, contrast, and visibility with less halo effect than other state-of-the-art methods both qualitatively and quantitatively in most cases.

Region search is widely used for object localisation in computer vision. After projecting the score of an image classifier into an image plane, region search aims to find regions that precisely localise desired objects. The recently proposed region search methods, such as efficient subwindow search and efficient region search, usually find regions with maximal score. For some classifiers and scenarios, the projected scores are nearly all positive or very noisy, then maximising the score of a region results in localising nearly the entire images as objects, or causes localisation results unstable. In this study, the authors observe that the projected scores with large magnitudes are mainly concentrated on or around objects. On the basis of this observation, they propose a region search method for object localisation, named level set maximum-weight connected subgraph (LS-MWCS). It localises objects by searching regions by graph mode-seeking rather than the maximal score. The score density by localised region can be controlled by a parameter flexibly. They also prove an interesting property of the proposed LS-MWCS, which guarantees that the region with desired density can be found. Moreover, the LS-MWCS can be efficiently solved by the belief propagation scheme. The effectiveness of the author's method is validated on the problem of weakly-supervised object localisation. Quantitative results on synthetic and real data demonstrate the superiorities of their method compared to other state-of-the-art methods.

In the single-image dehazing problem, it is critical that the transmission is accurately estimated. However, the extracted transmission in the dark channel model cannot effectively deal with the edge and the sky area because of the poor applicability of the dark channel prior to these areas. This study aims to solve that problem by proposing a novel variational model (VM) to optimise the transmission. This VM introduces a smoothness term and a gradient-preserving term to mitigate the false edge and the distorted sky area in the recovered image. Further, a fast algorithm to solve the VM is proposed on the basis of the additional operator splitting algorithm. This algorithm is an effective linear time algorithm and has excellent performance on optimising the transmission. The average running time of the algorithm shows an improvement of over 20 times that of the guided image filtering in these experiments. Experimental results also show that the proposed algorithm is both effective and efficient for optimising the transmission.

An efficient and accurate point matching algorithm named advanced weight graph transformation matching (AWGTM) is proposed in this study. Instead of relying only on the elimination of dubious matches, the method iteratively reserve correspondences which have a small angular distance between two nearest-neighbour graphs. The proposed algorithm is compared against weight graph transformation matching (WGTM) and graph transformation matching (GTM). Experimental results demonstrate the superior performance in eliminating outliers and reserving inliers of AWGTM algorithm under various conditions for images, such as duplication of patterns and non-rigid deformation of objects. An execution time comparison is also presented, where AWGTM shows the best results for high outlier rates.

Time-of-flight (TOF) sensors provide real-time depth information at high frame-rates. One issue with TOF sensors is the usual high level of noise (i.e. the depth measure's repeatability within a static setting). However, until now, TOF sensors’ noise has not been well studied. The authors show that the commonly agreed hypothesis that noise depends only on the amplitude information is not valid in practice. They empirically establish that the noise follows a signal-dependent Gaussian distribution and varies according to pixel position, depth and integration time. They thus consider all these factors to model noise in two new noise models. Both models are evaluated, compared and used in the two following applications: depth noise removal by depth filtering and uncertainty (repeatability) estimation in three-dimensional measurement.

Uncontrolled illumination is one of the most widely researched and most encountered face recognition challenges in recent years. In this study, the authors propose the division of algorithms into two categories: (i) relighting and (ii) unlighting. Relighting methods try to match the probe's illumination conditions using a subset of representative gallery images, while unlighting methods seek to suppress the variations. A total of 64 state-of-the-art methods are summarised and categorised in each of the groups. To make this work concise and easy to follow, they restricted themselves to selected conferences/journals and they limited the number of approaches reviewed. Also, eight past state-of-the-art approaches are used in both identification and verification experiments. However, only significant reported results from all methods were compared and organised in tables. The author's main objective is not to provide an exhaustive analysis of each category, but to present a collection of papers that can be useful in identifying research directions. Results indicate that unlighting methods are a better and a practical solution to address illumination challenges.