A new local feature descriptor recursive Daubechies pattern (RDbW) is developed by defining and encoding the Daubechies wavelet decomposed center–neighbour pixel relationship in the local texture. RDbW features are applied in spatial alignment (registration) of multimodal medical images using a Procrustes analysis (PA)-based affine transformation function and the registered images are further fused by employing a wavelet-based fusion method. A significant amount of experiments is conducted and the registration and fusion accuracy of the proposed feature descriptor is compared with the prominent existing methods such as local binary patterns (LBP), local tetra pattern (LTrP), local diagonal extrema pattern (LDEP), and local diagonal Laplacian pattern (LDLP). Experimental results show the present registration method improves the average registration accuracy by 38, 47, 71, and 76% in contrast to LDLP, LDEP, LTrP, and LBP, respectively. Further, the fusion results of the current approach exhibit an average improvement in entropy by 11%, standard deviation by 6% edge strength by 12%, sharpness by 23%, and average gradient by 16% when compared with all other feature descriptors used for registering the images. Concepts presented here can be used widely in analysing the combined information present in multimodal medical images.

Depth images extracted by light-coding-based depth cameras are widely used to reconstruct three-dimensional scenes in recent years. However, the retrieved depth accuracy greatly influences the reconstruction quality. Here, the authors present an appropriate depth extraction method based on subpixel matching to improve the depth accuracy. The proposed method utilises nearest neighbour interpolation to the projector's image plane to obtain depth values at subpixel accuracy, thereby better preserving the important depth information without changing any inner structure of the depth camera. Experimental results show that the proposed method improves the depth images both on image quality and depth accuracy.

Three-dimensional (3D) images have become increasingly popular in practice. They are commonly used in medical imaging applications. In such applications, it is often critical to compare two 3D images, or monitor a sequence of 3D images. To make the image comparison or image monitoring valid, the related 3D images should be geometrically aligned first, which is called image registration (IR). However, IR for 3D images would take much computing time, especially when a flexible method is considered, which does not impose any parametric form on the underlying geometric transformation. Here, the authors explore a fast-computing environment for 3D IR based on the distributed parallel computing. The selected 3D IR method is based on the Taylor's expansion and 3D local kernel smoothing. It is flexible, but involves much computation. The authors demonstrate that this fast-computing environment can effectively handle the computing problem while keeping the good properties of the 3D IR method. The method discussed here is therefore useful for applications involving big data.

Segmentation is an important step in iris recognition framework because the accuracy of the iris recognition system is affected by the segmentation of the iris. The image acquisition introduces noise artefacts such as specular reflections, eyelids/eyelashes occlusions and overlapping intensities, which makes the segmentation process difficult. An efficient method has been proposed for the segmentation of iris images that deal with non-circular iris boundaries and other noise artefacts mentioned above. The proposed method uses the Otsu multilevel thresholding based on improved particle swarm optimisation technique as a pre-segmentation step. Pre-segmentation step delimits the iris region from the other parts of an eye image. The geodesic active contours incorporated with a novel stopping function is then used to segment non-circular iris boundaries. The recognition accuracy of the proposed method is verified using the standard databases, CASIA v3 Interval and UBIRISv1. Obtained results have been compared with existing methods and have an encouraging performance.

Due to uneven illumination and dim environment in the tunnel, the monitored image is blurred, which makes it difficult to recognise the traffic status. Therefore, it is necessary to enhance the tunnel image in advance. In this study, a fast image enhancement algorithm based on imaging model constraint is proposed. First, the method uses the combination of global atmospheric light and partitioned atmospheric light to estimate the local atmospheric light. Second, the transmission is estimated based on the formula derived from the imaging model constraints. Third, the method uses a constant instead of illumination to balance tunnel image illumination. Last, the tunnel image is enhanced according to the imaging model. Experimental and comparative analysis results show that the proposed method can rapidly and effectively enhance the tunnel image.

Denoising of hyperspectral images is an essential step to remove the visual artifacts and improve the quality of an image. There are various sources of noise such as dark current, thermal and read noise produced due to detectors, stochastic error of photo-counting and so on which leads to variability of noise both in spatial and spectral domains. In this study, author proposes a novel denoising method based on concept of Hilbert vibration decomposition (HVD). Being iterative in nature it segregates initial amplitude composition into various components which are composed of slow varying wavelength. Any hyperspectral image is captured by the sensor over contiguous wavelengths. Thus, variation in intensities over the spectral dimension is less. HVD separates pixels in decreasing order of their intensity and results in denoising of the image. To evaluate method, various noise conditions have been tested on three real datasets: Washington DC mall, Urban and Pavia University. The validation is done both visually and quantitatively. The denoising with almost 100% mean structural similarity index confirms superiority of the designed method. Clustering and spectral analysis of various denoised images have also been reported. Clustering accuracy of 65% is achieved by the HVD as compared to other methods.

Non-rigid point set registration is a key component in many computer vision and pattern recognition tasks. In this study, the authors propose a robust non-rigid point set registration method based on high-dimensional representation. Their central idea is to map the point sets into a high-dimensional space by integrating the relative structure information into the coordinates of points. On the one hand, the point set registration is formulated as the estimation of a mixture of densities in high-dimensional space. On the other hand, the relative distances are used to compute the local features which assign the membership probabilities of the mixture model. The proposed model captures discriminative relative information and enables to preserve both global and local structures of the point set during matching. Extensive experiments on both synthesised and real data demonstrate that the proposed method outperforms the state-of-the-art methods under various types of distortions, especially for the deformation and rotation degradations.

This study presents a circle Hough transform (CHT) architecture that provides memory reduction between 74 and 93% without and with little degradation in the accuracy, respectively. For an image of P × Q pixels, the standard (direct) CHT requires a two-dimensional (2D) accumulator array of P × Q cells, but the proposed CHT uses a 2D accumulator array of (P/m) × (Q/n) cells for coarse circle detection and two 1D accumulator arrays of P × 1 and Q × 1 cells for fine detection, therein reducing the memory by a factor of m × n (approximately). The proposed CHT architecture was applied to iris localisation application and carried out its comprehensive evaluation. The average accuracy of the proposed CHT for iris localisation (inner plus outer iris-circle detection) is 98% with memory reduction of 87% compared with the direct CHT. The proposed CHT architecture was implemented on field programmable logic array targeting Xilinx Zynq device. The proposed CHT hardware takes processing time of 6.25 ms (average) for iris localisation in an image of 320 × 240 px². The proposed work is compared with the previous work, which shows improved results. Finally, the effect of additive Gaussian noise on the CHT performance is investigated.

A novel framework is developed for the modelling and clustering of proportional data (i.e. normalised histograms) based on the Beta-Liouville mixture model. This framework is based on incremental model selection, by testing if a given component was truly Beta-Liouville distributed. Specifically, the authors compare the theoretical maximum entropy of the given component with the estimated entropy obtained by the MeanNN estimator. If a significant difference was gained from this comparison, this component is considered as not well fitted and is then splitted into two new components with a proper initialisation. Our approach is tested through synthetic data sets and real-world applications which involve human gesture recognition and vehicle tracking for traffic monitoring purposes, which demonstrate that the authors' approach is superior to comparable techniques.

The optimisation of measurement matrix that is within the compressive sensing framework is considered in this study. Based on the fact that an information factor with smaller mutual coherence performs better, the gradient measurement matrix optimisation method is improved by an orthogonal search direction revision factor. This algorithm updates the approximation of ideal Gram matrix of information operator and the measurement matrix alternatingly. Using measurement matrix and sparse basis to represent the Gram matrix, the measurement matrix is optimised by the gradient algorithm, in which an orthogonal gradient search direction revision factor is proposed and utilised to further improve the performance of measurement matrix. This orthogonal factor is computed by the Cayley transform of a real skew symmetric matrix that is related to the gradient and the measurement matrix. Results of several experiments show that compared with the initial random matrix, the optimised measurement matrix can lead to better signal reconstruction quality.

This study demonstrates the development of vision based static hand gesture recognition system using web camera in real-time applications. The vision based static hand gesture recognition system is developed using the following steps: preprocessing, feature extraction and classification. The preprocessing stage consists of illumination compensation, segmentation, filtering, hand region detection and image resize. This study proposes a discrete wavelet transform (DWT) and Fisher ratio (F-ratio) based feature extraction technique to classify the hand gestures in an uncontrolled environment. This method is not only robust towards distortion and gesture vocabulary, but also invariant to translation and rotation of hand gestures. A linear support vector machine is used as a classifier to recognise the hand gestures. The performance of the proposed method is evaluated on two standard public datasets and one indigenously developed complex background dataset for recognition of hand gestures. All above three datasets are developed based on American Sign Language (ASL) hand alphabets. The experimental result is evaluated in terms of mean accuracy. Two possible real-time applications are conducted, one is for interpretation of ASL sign alphabets and another is for image browsing.

In this study, a block-based estimation of noise due to blurriness distortion is proposed based on auto-regressive (AR) modelling. In the proposed method; a de-correlated, low-energy version of the blurred image is auto regressively modelled. To this end, AR parameters are estimated using the Yule–Walker equations. As these equations include auto-correlation function (ACF) coefficients, ACF estimation is also required. The Yule–Walker equations are solved making use of Durbin–Levinson algorithm. Finally, noise energy is mathematically defined and computed for each block. Since blurriness is a signal-dependent image distortion, estimating and describing its characteristics via a noise like that of the AR model input, is significant. In fact, extracting features of such ‘noise’ can lead to the design and development of a new method of image quality metrics. Inspired by the ‘stem cells’ concept in medical science that is convertible to other cell types, the AR model input is called ‘stem noise’. To visualise contribution of the ‘Stem Noise’ in the reconstruction of blurriness image distortion, a map called stem noise energy map is created. It is shown that the characteristics of the estimated noise energy are well correlated with the human subjective scores.

To improve the image representation efficiency of trace transform (TT) features to images with circular and arc-shaped textures, the authors propose circular TT (CTT) to extract features. CTT consists of tracing an image with circles around which certain functionals of the image function are calculated. Quadruple CTT features can be generated through three successive functionals in the results of CTT, while different quadruple features can be obtained by choosing different combinations of successive functionals. These quadruple features can represent different texture properties and deeper intrinsic information of an image. By fusing CTT features and TT features based on PCA (FFCT_PCA), they construct a new complementary descriptor with much less dimension, further improving the representation performance for mixed texture images. Experimental results demonstrate that CTT has better performance than TT in recognising images with circular and arc-shaped textures, and FFCT_PCA has the potential to outperform the state-of-the-art feature extraction methods.

Even though face recognition is one of the most studied pattern recognition problems, most existing solutions still lack efficiency and high speed. Here, the authors present a new framework for face recognition which is efficient, fast, and robust against variations of illumination, expression, and pose. For feature extraction, the authors propose extracting Gabor features in order to be resilient to variations in illumination, facial expressions, and pose. In contrast to the related literature, the authors then apply supervised locality-preserving projections (SLPP) with heat kernel weights. The authors’ feature extraction approach achieves a higher recognition rate compared to both traditional unsupervised LPP and SLPP with constant weights. For classification, the authors use the recently proposed sparse representation-based classification (SRC). However, instead of performing SRC using the computationally expensive minimisation, the authors propose performing SRC using fast matching pursuit, which considerably improves the classification performance. The authors’ proposed framework achieves ∼99% recognition rate using four benchmark face databases, significantly faster than related frameworks.

To improve the poor robustness and low accuracy of the existing algorithms of image splicing detection, a novel passive image forgery detection method is proposed in this study, which is based on DOCT (discrete octonion cosine transform) and Markov. By introducing the octonion and DOCT, the colour information of six image channels (the RGB model and the HSI model) can be exhaustively extracted, which enhances the robustness of the algorithm. On the issue of improving the detection accuracy, the standard deviation is used to characterise the relationship of the colour information between the parts of DOCT coefficient matrix, and the K-fold cross-validation is introduced to improve the identification performance of the classifier. The steps of the algorithm are as follows: Firstly, the 8 × 8 block DOCT transform is used to the original image to obtain parts of block DOCT coefficient. Secondly, the standard deviation is used to process the corresponding parts of all blocks of the image. Finally, the Markov feature vector of the DOCT coefficient is extracted and feds to the LIBSVM (a library for support vector machines). When using LIBSVM for classification, K-fold cross-validation is executed to select the best parameter pairs. The experiment results demonstrate that the algorithm is superior to the other state-of-the-art splicing detection methods.

Hyperspectral imaging has been widely used for agriculture, astronomy, surveillance, and so on. However, hyperspectral imaging usually suffers from low-spatial resolution, due to the limited photons in individual bands. Recently, more hyperspectral image super-resolution methods have been developed by fusing the low-resolution hyperspectral image and high-resolution RGB image, but most of them did not consider the misalignment between two input images. In this study, the authors present an effective method to restore a high-resolution hyperspectral image from the misaligned low-resolution hyperspectral image and high-resolution RGB image, which exploits spectral and spatial correlation in hyperspectral and RGB images. Specifically, they employ the spectral sparsity to restore the high-resolution hyperspectral image on the misaligned part, and then simultaneously employ spectral and spatial structure correlation to restore the high-resolution hyperspectral image on the aligned area, which can be fused to obtain the high-quality hyperspectral image restoration under a misaligned hybrid camera system. Experimental results show that the proposed method outperforms the state-of-the-art hyperspectral image super-resolution methods under a misaligned hybrid camera system in terms of both objective metric and subjective visual quality.

Background subtraction (BS) is a fundamental step for moving object detection in various video surveillance applications. Gaussian mixture model (GMM) is a widely used BS technique which provides a good compromise between robustness to the background variations and real-time constraints. However, GMM does not support the spatial relationship among neighbouring pixels and it uses a fixed learning rate for every pixel during the parameter update. On the other hand, Wronskian change detection model (WM) is a spatial-domain BS technique which solves misclassification of pixels but fails in the presence of dynamic background. In this study, a novel spatio-temporal BS technique is proposed that exploits spatial relation of Wronskian function and employs it with a new fuzzy adaptive learning rate in a GMM framework. Instead of using WM directly, an improved WM is proposed by adaptively finding out the ratio of the current pixel to the background pixel or its reciprocal, and a weighted Wronskian is developed to mitigate the effect of dynamic background pixels. Additionally, a new fuzzy adaptive learning rate is employed in the GMM framework. Experimental results of the proposed framework yield better silhouette of the moving objects as compared with the state-of-the-art techniques.

Here, the author presents the gamma binary distance, an exceptional distance for measuring similarity between binary images. The gamma distance is a probabilistic pixel mapping measure that is a modification of the Hamming distance. Employing a probabilistic approach to image matching enables gamma to measure similarity more accurately than employing traditional binary distances. The author shows the advantage of employing the gamma distance for similarity measurement by comparing it to three of the most popular similarity distances used for binary image matching: correlation, sum of the absolute difference method, and mutual information. Results of extensive testing conducted on a large database are presented where the superiority of the gamma distance over other similarity distances is shown.

Biometrics authentication is considered as most secure and reliable method to recognise and identify person's identity. Researchers put efforts to find efficient ways to secure and classify the solutions to biometric problems. In this category, fingerprint recognition (FPR) is most widely used biometric trait for person identification/verification. The present work focuses an FPR technique, which uses the grey-level difference method, discrete wavelet transforms and edge histogram descriptor for fingerprint representation and matching. Wavelet shrinkage used for noise removal in the image. Ridge flow estimation is calculated using the gradient approach. SVM and Hamming distance similarity measures are used for recognition. The experiment result has been tested on the standard 2000–2004 fingerprint verification competition dataset and the accuracy of proposed algorithm was reported to be well above 98%.

Rain image enhancement is important for outdoor computer vision applications. In this study, the authors propose a multi-stage filtering method for single rainy image enhancement. It is based on their new rainy image model, and consists of two main operations: rain streaks removal and rain fog removal. For rain streaks removal, based on one key observation that the low-pass version of a rainy image and that of a non-rainy image of the same scene are almost the same after appropriate low-pass filtering, they remove rain streaks from rainy images by decomposing an input rainy image (or a rainy component image) into the low-frequency (LF) part and the high-frequency (HF) part via an LF smooth filter, i.e. the traditional Gaussian filter with a simple subtraction operation in multiple different stages. After rain streaks removal, dark channel prior-based method was employed for rain fog removal. Experimental results show that the proposed algorithm generated comparable outputs with most of the state-of-the-art algorithms with low computation cost.

Magnetic resonance imaging (MRI) and positron emission tomography (PET) image fusion is a recent hybrid modality used in several oncology applications. The MRI image shows the brain tissue anatomy and does not contain any functional information, while the PET image indicates the brain function and has a low spatial resolution. A perfect MRI–PET fusion method preserves the functional information of the PET image and adds spatial characteristics of the MRI image with the less possible spatial distortion. In this context, the authors propose an efficient MRI–PET image fusion approach based on non-subsampled shearlet transform (NSST) and simplified pulse-coupled neural network model (S-PCNN). First, the PET image is transformed to YIQ independent components. Then, the source registered MRI image and the Y-component of PET image are decomposed into low-frequency (LF) and high-frequency (HF) subbands using NSST. LF coefficients are fused using weight region standard deviation (SD) and local energy, while HF coefficients are combined based on S-PCCN which is motivated by an adaptive-linking strength coefficient. Finally, inverse NSST and inverse YIQ are applied to get the fused image. Experimental results demonstrate that the proposed method has a better performance than other current approaches in terms of fusion mutual information, entropy, SD, fusion quality, and spatial frequency.

This study presents a geometric deformable model-based segmentation approach to segmentation of the intima and media-adventitial (MA) borders in sequential intravascular ultrasound (IVUS) images. The initial estimation of the vessel borders was done manually only for the first frame of each sequence. After the border initialisation, pre-processing including edge preservation, noise reduction, and dead zone preservation was successively performed on each IVUS frame. To improve segmentation performance, the image masks were determined preliminarily by local binary pattern-based mask initialisation. Then, the inner and outer borders were approximated using a modified distance regularised level set evolution model. The results showed superior performance of the suggested approach for estimating intima and MA layers from the IVUS images. The corresponding correlation coefficients of area, vessel perimeter, maximum vessel diameter, and maximum lumen diameter were r = 0.782, r = 0.716, r = 0.956, and r = 0.874 for the 20 MHz images, respectively, and r = 0.990, r = 0.995, r = 0.989, and r = 0.996 for the 45 MHz images, respectively. In addition, linear regression analysis indicated that the manual segmentation had significantly high similarity at r > 0.967 and r > 0.993 for 20 and 45 MHz images, respectively.

L0 gradient minimisation model, one of edge-aware image smoothing method, also suffers fromthe stair-casing effect and images with strong textures cannot be smoothed effectively and weak edges or structureswill be smoothed overly. The authors propose a method to overcome these drawbacks above. To begin with, theimage is subjected to non-subsampled shearlet transform to obtain high-frequency component, and combine allhigh-frequency component by maximum local energy rules to obtain the high-frequency decomposition image,afterwards, introducing the data term associated with high-frequency decomposition image to keep the similarityof edge and structure between the input and smoothed image. Secondly, the patched L0 gradient minimisationmodel is presented for improving the description of local information, since different size of the patches has thedifferent texture, exploiting the coefficient of variation to define the size of patch. Finally, defining the adaptivesmoothing coefficient based on the gradient to make sure that the smoothing effect of the patch is optimal. Theproposed model is applied to image smoothing with desirable results successfully, and the comparisons with otherstate-of-the-art edge-preserving image smoothing algorithms demonstrate the great performances ofedge-preserving and texture smoothing.

The spontaneous proliferation of video data necessitates implementing hierarchical structures for various content management applications. Temporal video segmentation is the key towards such management. To address the problem of temporal segmentation, the current communication exploits the concept of psychological behaviour of the human visual system. Towards this goal an abrupt cut detection scheme has been proposed based on Weber's law which provides a strong spatial correlation among the neighbouring pixels. Thus, the authors provide a robust and unique solution for abrupt shot boundary detection when the frames are affected partially or fully by flashlight, fire and flicker, high motion associated with an object or camera. Further, they have devised a model for generating an automatic threshold, taking into account the statistics of the feature vector which quadrates itself with the variation in the contents of the video. The effectiveness of the proposed framework is validated by exhaustive comparison with few contemporary and recent approaches by using benchmark datasets TRECVID 2001, TRECVID 2002, TRECVID 2007 and some publicly available videos. The results obtained give credence to the remarkable improvement in the performance while preserving a good trade-off between missed hits and false hits as compared to the state-of-the-art methods.

This study introduces an encryption algorithm for colour RGB images and text, the encryption is based on the synchronisation of fractional chaotic systems, the synchronisation has the topology of master–slave, where the transmitter is the master system and the receiver is the slave system, this last one is designed as a new smoothed sliding modes state observer for fractional chaotic systems. The encryption algorithm provides security against common cryptographic techniques, including known and chosen plaintext attacks.