The challenge of visual tracking mainly comes from intrinsic appearance variations of the target and extrinsic environment changes around the target in a long duration, so the tracker that can simultaneously tolerate these variabilities is largely expected. In this study, the authors propose a new tracking approach based on discriminative stable regions (DSRs). The DSRs are obtained based on the criterion of maximal local entropy and spatial discrimination, which enables the tracker to handle well distractors and appearance variations. The collaborative tracking incorporated hierarchical clustering can tolerate motion noise and occlusions. In addition, as an efficient tool, the principle component analysis is used to discover the potential affine relation between DSR and the target, which timely adapts to the shape deformation of the target. Extensive experiments show that the proposed method achieves superior performance in many challenging target tracking tasks.

A new method for image thresholding of two or more images that are acquired in different modalities or acquisition protocols is proposed. The method is based on measures from information theory and has no underlying free parameters nor does it require training or calibration. The method is based on finding an optimal set of global thresholds, one for each image, by maximising the mutual information above the thresholds while minimising the mutual information below the thresholds. Although some assumptions on the nature of images are made, no assumptions are made by the method on the intensity distributions or on the shape of the image histograms. The effectiveness of the method is demonstrated both on synthetic images and medical images from clinical practice. It is then compared against three other thresholding methods

The authors propose a novel spatiotemporal constraint based on shape and appearance and combine it with a level-set deformable model for left ventricle (LV) segmentation in four-dimensional gated cardiac SPECT, particularly in the presence of perfusion defects. The model incorporates appearance and shape information into a ‘soft-to-hard’ probabilistic constraint, and utilises spatiotemporal regularisation via a maximum a posteriori framework. This constraint force allows more flexibility than the rigid forces of shape constraint-only schemes, as well as other state of the art joint shape and appearance constraints. The combined model can hypothesise defective LV borders based on prior knowledge. The authors present comparative results to illustrate the improvement gain. A brief defect detection example is finally presented as an application of the proposed method.

Eye state analysis (open or closed) is an important step in fatigue detection. In this study, an efficient algorithm for eye state detection is proposed. At first, a new face detection method is presented for noisy images that finds the face area in the input image well. Then, novel algorithms for detection of eye region and eye state are introduced. The performance of the proposed method is evaluated on four different databases namely FERET, Aberdeen, IMM and CVL which contain more than 5700 images with different descents, positions, light conditions and glasses. The results show that the new method achieves more accuracy rate than the previously presented algorithms, while it does not need training data and is also computationally efficient.

The proposed method is based on extraction of non-linearity from the nearest neighbourhood elements of image. To detect non-linearity, relation between the nearest neighbourhood elements of the image, have been expressed in terms of Gaussian membership functions. All the elements of the image are connected with the nearest neighbourhood elements with some membership degree of the Gaussian functions. It results in the formation of number of fuzzy lattices. The lattices have been expressed in the form of Schrödinger equation, to find the kinetic energy (KE) used, corresponding to change occurring in the facial activity of a person. Finally, the KE embedded in three dimension space is used to distinguish non-linear changes during occurrence of various facial activities. Experimental results show that proposed method is effective in recognition of facial expression as it focuses on extracting the non-linear features corresponding to contours of maximum energy which are appearing because of different expressions.

Mobile mapping systems that detect and geo-reference road markings almost always consist of a stereo camera system integrated with a global positioning system/inertial navigation system. The data acquired by this navigational system allows features detected in the stereo images to be assigned global co-ordinates. An essential step in this process is the calibration of the cameras, as it relates the pose of the two cameras to each other and a world co-ordinate system. In Europe, road markings must be evaluated from a 35 m range, so the cameras are required to have a wide field of view. Traditional calibration methods supposedly require a calibration object that would fill most of the calibration images. This large field of view would require a calibration object of substantial size that would be impractical for the purposes of this portable system. This study explores the theory of camera calibration and then details two camera calibration techniques (using portable 3D and 2D calibration objects). The accuracy of these methods is then evaluated using a ground-truth experiment.

A novel approach is presented to gain high classification rate for each class of ImageCLEF 2007 medical database. The learning phase consists of four iterations where different classification models were generated as per iteration. For the iterations, a model generation process was performed in two steps. The first step starts with construction of a model from the entire dataset. This model was then assessed to filter high accuracy classes (HAC). These classes were those predicted with an accuracy rate above 80%. This evaluation performed on 20% of the training dataset was taken as test data. In the second step, classes under HAC were only used to construct the classification model. The same processes will be performed in the next iteration on the classes which were left with accuracy below 80% from the previous iteration. The methodology presented is based on a bag of visual words for feature extraction and the radial basis function (RBF)-based support vector machine classifier. As a result, four classification models were generated from 77, 17, 12 and 10 classes, respectively. These models were constructed and evaluated on a database consisting of 11 000 medical X-ray images (training dataset) and 1000 (testing dataset) of 116 classes. The accuracy rate obtained by each generated model outperformed the results obtained by only one model on the entire dataset.

An adaptive shadow detection algorithm is proposed to eliminate interference on object detection from the shadow. The algorithm uses three components in YUV colour space to identify shadow pixels from the candidate foreground. An adaptive threshold estimator is designed to improve shadow detection accuracy and adaptive capacity in various lighting conditions. This estimator uses edge detection method to obtain global texture, as well statistical calculations to obtain the thresholds. Algorithm has the characteristic of low complexity and little restraint; hence it is suitable for real time-moving shadow detection in various lighting conditions. Experiment results show that this algorithm can obtain a high detection accuracy and the time-assume is greatly shortened compared with other algorithms with similar accuracy.

In this study, the authors present a new area-based stereo matching algorithm that computes dense disparity maps for a real-time vision system. Although many stereo matching algorithms have been proposed in recent years, correlation-based algorithms still have an edge because of speed and less memory requirements. The selection of appropriate shape and size of the matching window is a difficult problem for correlation-based algorithms. In the proposed approach, two correlation windows are used to improve the performance of the algorithm while maintaining its real-time suitability. The CPU implementation of the proposed algorithm computes more than 10 frame/s. Unlike other area-based stereo matching algorithms, this method works very well at disparity boundaries as well as in low textured image areas and computes a dense and sharp disparity map. Evaluations on the benchmark Middlebury stereo datasets have been performed to demonstrate the qualitative and quantitative performance of the proposed algorithm.

Laser range finder and omnidirectional cameras are becoming a promising combination of sensors to extract rich environmental information. This information includes textured plane extraction, vanishing points, catadioptric projection of vertical and horizontal lines, or invariant image features. However, many indoor scenes do not have enough texture information to describe the environment. In these situations, vertical edges could be used instead. This study presents a sensor model that is able to extract three-dimensional position of vertical edges from a range-augmented omnidirectional vision sensor. Using the unified spherical model for central catadioptric sensors and the proposed sensor model, the vertical edges are locally projected, improving the data association for mapping and localisation. The proposed sensor model was tested using the FastSLAM algorithm to solve the simultaneous localisation and mapping problem in indoor environments. Real-world qualitative and quantitative experiments are presented to validate the proposed approach using a Pioneer-3DX mobile robot equipped with a URG-04LX laser range finder and an omnidirectional camera with parabolic mirror.