As vehicular traffic continues to grow traffic management and prevention of accidents has become a major concern. This problem only gets magnified when travel on the mountainous roads are considered. This study is especially focused toward Himalayan mountains as they pose a greater risk because of their rugged natural setting. This study investigates crucial problems faced on the hilly roads and the challenges in translating existing driver-assistance systems to such roads. The survey probes every lane detection algorithms, image processing techniques and various assistance features for applicability to hilly roads discussing the pros and cons for each of them. Conclusions are drawn as to deduce the more suitable methods that can be improvised and re-tuned to adapt them for mountainous roads.

It is difficult to preserve diminishing weak structures and edges, and remove complex details simultaneously in the context of image smoothing. While most of existing methods only take either local or global features into consideration, the authors propose two methods taking advantage of both to achieve smoothing, both of which consist of two steps and share the same first step. In the first step, the authors use a scale-aware approach to generate a guidance image by blurring the small-scale components in the input image. Such approach, based on the rolling guidance framework with domain transform filter and bilateral filter, can prevent diminishing the corners of the main structures. Subsequently, the authors use the two proposed methods, with the guidance image as input, to remove blurry details. The first method introduces two data fidelity terms into L ₀ gradient minimisation and removes high-contrast details, which is a structure-preserving method. The other method, an edge-preserving method, uses an adaptive L ₀ gradient minimisation technique, facilitating the preservation of the weak structures and edges. The smoothing factors in such technique are decide by the corresponding gradient of each pixel of the guidance image. The authors apply both methods to various image processing fields.

A new anti-occlusion particle filter object-tracking method based on feature fusion is proposed in this study. Colour and local binary pattern features are extracted and additively fused with a deterministic coefficient, which is calculated based on the difference between the object features and the background. An integral cumulative histogram is proposed to reduce the computational cost of feature extraction. A new occlusion determination method is proposed, and corresponding tracking strategies are also put forward for various occlusion conditions; in the case of partial occlusion, block tracking is carried out, and in the case of serious occlusion, the least-square method is used to predict the object position. Context Aware Vision using Image-based Active Recognition (CAVIAR) and Video Image Retrieval and Analysis Tool (VIRAT) video libraries are used to validate the method. The experimental results show that the proposed method can describe an object effectively and improve tracking stability and robustness under the occlusion conditions.

Traditional tomography uses circular trajectories and here, filtered backprojection often works well. However, for objects with large aspect ratios, rotational tomography is often not feasible. In these cases, other trajectories can be more appropriate. For generic trajectories, filtered backprojection methods might not work well and full iterative reconstruction can be computationally demanding. In this study, the authors thus propose a third paradigm that combines aspects of both of these techniques. They use interpolation and backprojection techniques to generate an initial estimate of an object's internal structure using projection images taken at different orientations. Depending on the scanning geometry used to calculate the tomographic projections, this initial estimate can be understood as a blurred (filtered) approximation of the actual structure. For each scanning geometry, they specify the equivalent blurring operator that would provide the same estimate directly from a representation of the object's internal structure. They then use iterative techniques to invert this filtering operation, thus estimating the internal structure from the estimate of its blurred representation.

Advances in the eye care telemedicine system aid the diabetic patients in remote areas to stop the unwanted visit to ophthalmologist, reduces overall cost, time and money. Diabetic retinopathy, which is the primary cause of sight loss, has the most common symptoms like microaneurysms, haemorrhages, cotton-wool spots, exudates and drusen. In this work, an efficient approach for the automatic detection of haemorrhages in colour retinal images is proposed and validated. The colour retinal images captured from the diabetic patients are enhanced by an effective pre-processor. A bag of features based on intensity, colour and texture are extracted. Finally, the features are classified with the help of partial least square classifier. The classifier performance is validated on two publicly available databanks. The developed method obtains an area under receiver operating characteristic curve of 0.98 with an average execution time of 6 s. This application outperforms the existing approaches with high robustness and efficiency.

The methods of the image noise reduction based on optimal channel-processing to enhance image quality are studied. According to different noise with the different ratio in colour components, different noise reduction technologies are used for noise reduction. Then the optimal algorithm is automatically selected as the ultimate way of noise reduction in each channel. For the purpose of optimal noise reduction effect, this study presents a method of combining the quadratic optimisation with the variable window processing. The quadratic optimisation provides a good environment for noise reduction by decreasing complexity of mixed noise and the variable window processing calibrates the image smoothing result. Compared with mean filtering, median filtering and adaptive filtering, the image quality processed by the proposed algorithm is generally improved by >2 dB.

To solve problems that colour distortion and low resolution of infrared and visible colour fusion images, the authors propose a fusion method based on the double-opponency colour constancy mechanism of human vision. First, Waxman's fusion method which imitated the neuro-dynamics mechanism of the rattlesnake bimodal cell is used to fuse the visible light with the source multi-band images to generate pseudo-colour images. Second, a double-opponent colour constancy computation model based on Rodieck's double difference-of-Gaussian is proposed to obtain the estimate of an illuminant of colour fusion images. Finally, the colour fusion images are corrected by the diagonal transformation model based on the cone adaptive mechanism. They also propose to use the three-dimensional RGB histogram to analyse the colour distribution of colour fusion images. In the comparison experiments with other approaches using the three-dimensional RGB histogram, one can see that the proposed fusion method gives colour image coinciding well with natural colour distribution and satisfies human perception needs.

Motion estimation is one of the most crucial and time-consuming component of video compression methods. However, much research has been done to improve computational complexity at the expense of the loss in performance of matching of blocks. A novel block matching algorithm named efficient direction-oriented search is proposed. For this, the proposed algorithm firstly aims to dynamically switch between search regions based on the location of minimum distortion error. The search region dimension is also made adaptive for faster convergence. Then the computational complexity is reduced by using a proposed horizontal, vertical wings diamond search pattern and, two inclined hexagon-shaped direction-oriented search patterns. For further speed-up in the search process, partial distortion calculations are employed. A method for optimal threshold value selection based on the distortion statistics for different partial distortion calculations is presented. The performance of the proposed algorithm is evaluated for different video sequences containing: slow, medium, fast, and directional motion content. The experimental results indicate that significant improvement in speed-up can be achieved while maintaining the better peak signal-to-noise-ratio performance. For directional motion video sequences, the proposed method even outperforms the full search algorithm with a significantly lower computational cost.

Restoring pixel intensities corrupted by impulse noise has a great impact on the quality of decision-based filters. In this study, the authors’ focus is on intensity restoration of noisy pixels. Their assumption is that noisy pixels are already established by the noise-detection unit being considered as missing data in the image. When the interpolation methods are adopted in the noise-restoration unit of the decision-based filters for the purpose of restoring the intensities of the noisy pixels, two unexpected problems emerge – jagged edges and blurred details. These drawbacks can be ameliorated by using extra information obtained from structures in the images. Their structure-based interpolation method comprises two steps: pre-interpolation and post-interpolation. In the first step (pre-interpolation), the Sibson natural neighbour interpolation is adopted for the initial estimation of the intensities of all noisy pixels. In the second step (post-interpolation, modifying-phase), for each noisy pixel in pre-interpolated image, the intensity variations of the pixels on two adjacent parallel lines, in different directions in their corresponding local windows, are analysed. Based on the obtained structural information, the intensity of the centred noisy pixel is restored more effectively. Since the structures in the images are far more noticeable at low-density impulse noise, the proposed method works more efficiently in this case; however, a gradual improvement is achieved for high-density impulse noise.

Recently, kernelised correlation filter (KCF)-based trackers aroused increasing interest and achieved extremely compelling results in different competitions and benchmarks in the field of visual object tracking. However, the training mechanism of the KCF that exploits simple linear combinations of filter from the previous frame easily cause error accumulation. To overcome this problem, the authors propose a novel training strategy that utilises all of the previous training samples, and a sparsity-related loss function regularised by the L1 norm to deal with the problem of the fixed template size in KCF trackers, a separate scale filter is learned for scale estimation during the tracking process. Moreover, powerful features that include histogram of oriented gradients (HOG) and colour features are integrated to further improve the robustness of the authors’ tracking. Extensive experiments in various challenging situations demonstrate that the proposed method performs favourably against several state-of-the-art tracking algorithms.

In this paper, a new technique for plane detection from 3D point clouds is proposed. The algorithm depends on two concepts to balance between high-accuracy and fast performance. The first is the use of a new fast octree-based balanced density down-sampling technique to reduce the number of points. The second is the fact that the number of planes in any dataset is much less than the number of the points. Random points are examined to find the 3D planes. To increase the accuracy, the system utilizes an adaptive plane extraction technique to overcome data noise. Initially, the point cloud is subdivided using octree into small cubes with a limited number of points. Then the cubes are down-sampled based on the local density of each cube. The points are explored randomly for finding a planar surface by applying principal component analysis (PCA) on the points’ spherical neighborhood obtained by the down-sampled octree structure. The adaptive plane extraction is used to adjust the plane orientation to find the best position that includes the maximum number of points. Experimental results demonstrate that the proposed algorithm is capable of processing large amounts of data efficiently to produce accurate results that are robust to noise.

In this study, the authors present a new feature extraction method for handwritten Arabic/Persian language word recognition. This feature is based on the angle, number, location, and size of straight lines which represents geometric and quantitative attributes of a word. At first, word image is broken into an m × n window and straight lines are extracted from each window. Then, the proposed features are taken from these lines and combined together. Finally, the features of the images are used for training and testing support vector machine classifier. The proposed method is tested on three datasets: IBN-SINA and IFN/ENIT for Arabic words and Iran-cities for Persian words recognition. Recognition accuracy of the proposed method is about 67.47, 86.22 and 80.78% for the Iran-cities, IBN-SINA and IFN/ENIT Arabic dataset, respectively, which is better than state-of-the-art methods.

An adaptive plateau limit-based histogram equalisation algorithm is suggested to enhance digital images. Histogram of the image is clipped with a plateau limit to avoid over enhancement. The plateau limit is derived from the average of the mean and the median intensity values to offer the improved enhancement. Clipped histogram is subdivided into three parts, using histogram subdivision limit parameters that are calculated on the basis of the standard deviation of the image. Histogram of individual sub-image is equalised independently and then combined into a single enhanced image. Experimental results demonstrate that the proposed plateau limit-based tri-histogram equalisation algorithm enhances the image quality. Compared with the other traditional plateau and non-plateau limit-based histogram equalisation algorithms, quantitative and visual quality assessments effectively validate the superiority of the proposed algorithm.

Wavelet-based segmentation approaches are widely used for texture segmentation purposes because of their ability to characterise different textures. In this study, the authors assess the influence of the chosen wavelet and propose to use the recently introduced empirical wavelets. We show that the adaptability of the empirical wavelet permits to reach better results than classic wavelets. To focus only on the textural information, they also propose to perform a cartoon + texture decomposition step before applying the segmentation algorithm. The proposed method is tested on six classic benchmarks, based on several popular texture images.

In this study, the authors present a simple, reliable, fast, unrestricted-shape geometry, and accurate algorithm which runs in O(log₂  n) time to find the axis-parallel largest rectangle (LR) inside a given region of interest (ROI), where n is the image size in one dimension, which means that the proposed model can work in real time. The proposed approach is successful in detecting the LR of arbitrary orientation that is fully contained in the ROI as well. Also, the present algorithm can find the largest empty rectangle in a space containing a set of zero points, whether the axis-parallel rectangle or the oriented one. The strategy followed here is to accelerate LR detection process by searching the rectangle with the largest area inscribed in the ROI, by starting first with the lowest-resolution version of the original image for determining the LR four corners’ coordinates, then next searching the new LR corners’ scaled coordinates in the higher power resolutions in a multiple resolutions hierarchical model and therefore, a corresponding coarse-to-fine inference procedure recursively eliminates the search space of the LR four corners coordinates. For finding the largest oriented rectangle, the same hierarchical procedures are followed, but combined with rotation-angle resolution.

The background subtraction is an important technique in computer vision which segments moving objects into video sequences by comparing each new frame with a learned background model. In this work, the authors propose a novel background subtraction method based on Gaussian–Bernoulli restricted Boltzmann machines (GRBMs). The GRBM is different from the ordinary restricted Boltzmann machine (RBM) by using real numbers as inputs, resulting in a constrained mixture of Gaussians, which is one of the most widely used techniques to solve the background subtraction problem. The GRBM makes it easy to learn the variance of pixel values and takes the advantage of the generative model paradigm of the RBM. They present a simple technique to reconstruct the learned background model from a given input frame and to extract the foreground from the background using the variance learned for each pixel. Furthermore, they demonstrate the effectiveness of the proposed technique with extensive experimentation and quantitative evaluation on several commonly used public data sets for background subtraction.

An image segmentation method based on a non-parametric mixture model together with total variation (TV) regularisation is proposed. The authors use a kernel density estimator as a basic mixture model, which can better separate the non-central distributed data. To enforce its robustness, they integrate the well-known TV regularisation into the statistical method. They use the dual method to efficiently solve the TV-related energy and get a new dual expectation maximisation algorithm. Experiments on both synthetic images and real images show that the proposed algorithm can achieve good segmentation results. Compared with the parametric models and hidden Markov random field-based method, the proposed method can produce better result in some cases.

Visual saliency is defined by the perceptual information that makes possible to detect specific areas which attract to guide the human visual attention. In this study, the authors present an efficient method for salient regions detection on three-dimensional (3D) meshes using weighted graphs representation. To do so, the authors propose a novel 3D surface descriptor based on a local homogeneity measure. Then, they define the similarity measure between vertices using normal deviation similarities, a two-dimensional projection height map, and the mean curvature. The saliency of a vertex is then evaluated as its degree measure based on the local patch descriptor and a height map. In addition, the authors introduce a custom version of hill climbing algorithm in order to segment the 3D mesh regions according to the saliency degree. Furthermore, they show the robustness of their proposed method through different experimental results. Finally, the authors present the stability and robustness of their method with respect to noise.

In computer vision, blob detection is used to obtain regions of interest that could signal the presence of objects or parts with application to object recognition and object tracking. One of the more common blob detectors is based on the Laplacian of Gaussian (LoG). However, most blob detectors developed in the past assume circular blobs, and these detectors do not perform as well with elliptical blobs, a more prevalent scenario in real images. A generalised LoG (GLoG) detector was proposed recently to deal specifically with elliptical blobs. To formulate the GLoG in a multi-scale framework, its response must be made scale invariant. Toward that end, necessary and sufficient conditions are presented here, with the normalisation factors derived for a scale-invariant GLoG detector. The factors are validated with a synthetic example and are further tested with two real-world images.

Here, the authors present an effective regularisation approach to colour image demosaicking. The authors’ work is inspired by the interesting observation that the cross-channel dependencies of high-frequency details of a colour image are supposed to correspond in all the main colour channels acquired by the sensor. Therefore, minimising their difference in the demosaicking process can significantly improve the quality of the reconstructed image. The authors also demonstrate that mosaicked image formation strictly depends on the intrinsic lens blur. Hence, in the authors’ solution to the image demosaicking as an inverse imaging problem, they take the lens blur characteristics into account. The proposed regularisation method is also based on the fact that sensor saturation significantly alters the distribution of pixel intensity and Gaussian noise. The authors develop an efficient solution to the problem via the alternating direction method of multipliers numerical solver. As a result of these steps, the proposed demosaicking approach significantly enhances the quality of reconstructed images. Experimental results and quantitative evaluations demonstrate that the proposed method outperforms the existing image demosaicking methods.