Colourisation is a process of adding colour to greyscale images. In this study, the authors propose two new colourisation models based on non-local total variation regularisation in the chromaticity and brightness (CB) colour space and the YIQ colour space. Lagrange multiplier method is used to handle the sphere constraint of chromaticity in the CB colour space. By introducing an extra variable and using the dual version of non-local total variation, they split the proposed colourisation problems into two subproblems with closed-form solutions and get two iterative algorithms. Experimental results and comparisons demonstrate that the advantage of the proposed methods is that they can preserve the colour edges better than the closely related existing methods, especially the total variation methods.

Most applications in computer vision manage to suppress textures and noise while maintaining meaningful structure based on colour intensity variation, but it is intractable due to texture patterns or error. This study presents an edge-preserving suppression method for depth estimation. The authors formulate a functional energy function based on the relative total intensity and space variation, and they minimise the energy function via iteratively reweighted least squares. Assuming that textural edges most likely correspond to depth discontinuities, they exploit the comparative variations of the colour image to produce a more accurate depth map. The experimental results demonstrate the usefulness of the proposed approach, and show that texture patterns are suppressed while meaningful edges are preserved. According to the results of the depth acquisition methods, the proposed depth estimation methods generate the accurate and robust results.

The authors propose a new single image dehazing method. Different from image restoration and image enhancement method, their method is based on the idea of image fusion. Image dehazing is to remove the influence of the haze between the scene and the camera. First, combined with the depth information, the haze layer is subtracted in the hazy image to improve the colour saturation, which produces the first input image. Then, the gamma correction is used on the grey image. Second, the details of the gamma correction image are enhanced to produce the second input image. Finally, the two input images are fused by local linear model to obtain the final restored image. Experimental results show that the restored image has high contrast, rich details, and without colour distortion in the sky area.

High-efficiency video coding (HEVC) is an international video coding standard. It can reduce bit rate by almost 50% when preceding state-of-the-art standard H.264/advanced video coding (AVC) with the same objective quality but increase about 40% encoding complexity. The computation complexity of HEVC largely relies on the inter-prediction, which consumes about 60–70% of the total encoding time. This study proposes a fast mode decision method for HEVC inter-prediction. Since current coding unit (CU) is the basic coding zone for HEVC encoding process, the authors apply the edge detection method with Sobel operator to extract the edge information inside CU. It is based on the threshold-oriented means to compare those edge values and then predict the best partition mode for this CU. Furthermore, they define a weighting factor to adopt various resolutions of test sequence from wide quarter video graphics array (WQVGA) to 1600 p. The experimental results show that the proposed algorithm can reduce computational complexity by 50% on average compared with HEVC reference software, and appear to have almost the same coding performance.

The major drawback of the fuzzy c-means (FCM) algorithm is its sensitivity to noise. The authors propose a new extended FCM algorithm based a non-parametric Bayesian estimation in the wavelet transform domain for segmenting noisy MR brain images. They use the Bayesian estimator to process the noisy wavelet coefficients. Before segmentation based on FCM algorithm, they use an a priori statistical model adapted to the modelisation of the wavelet coefficients of a noisy image. The main objective of this wavelet-based Bayesian statistical estimation is to recover a good quality image, from a noisy image of poor quality. Experimental results on simulated and real magnetic resonance imaging brain images show that their proposed method solves the problem of sensitivity to noise and offers a very good performance that outperforms some FCM-based algorithms.

This study proposes a local similarity refinement strategy to handle the parallax problem in image stitching. The proposed method is combined with deconvolution to acquire high-accuracy matching between corresponding source images. Shape-preserving half-projective warp was used to eliminate distortion across the non-overlapping region caused by the global projective transformation. The proposed refinement method further refines the warping result within the overlapping region, where it suppresses the parallax. The method was compared with various state-of-the-art methods: projective (global homography), AutoStitch, Zaragoza's method, Zhang's method, and Chang's approach. All comparisons are based on both public data sets and a proposed Inha University Computer Vision Lab (ICVL) stitching data set. The experimental results demonstrate that the proposed method is robust for handling the parallax in image stitching.

Automatic extraction of retinal blood vessels plays an important role in the diagnosis of many retinal diseases and also for diagnosing several complicated diseases such as stroke, hypertension and cardiovascular diseases. Due to the complex nature of retinal vessel network, the manual segmentation of vessels is a tedious task which also requires high training and skills. This study presents a new method for blood vessel segmentation in colour retinal images using supervised approach. Initially, a set of core features including Gabor filter responses, Frangi's vesselness measure (1D), local binary pattern feature (1D), Hu moment invariants (7D) and grey-level co-occurrence matrix features (3D) are considered. The neural network is trained with the different subsets of core features and it is found that the model with 13D features excluding the Hu moment invariants results in better performance. This model is used for evaluation. The proposed supervised segmentation approach is tested on publicly available structured analysis of the retina, digital retinal images for vessel extraction and CHASE_DB1 databases which contain manually labelled images. The performance of the proposed algorithm is evaluated on the basis of accuracy, sensitivity, specificity and area under the curve. The proposed technique achieves high mean accuracy and sensitivity while it is compared with the several previously proposed algorithms.

In this study, a novel normal curvature-induced variational model which involves a higher-order regulariser based on the normal curvature prior information of image surface is proposed for image restoration. Furthermore, the authors derive a preferably equivalent formulation for the proposed normal curvature-induced higher-order regulariser. Then, they design an efficient algorithm to solve the proposed model by using the famous alternating direction method of multipliers technique. Finally, they assess the performance of the proposed method on both natural images and biomedical cell images by comparing it with the famous fast total variation (TV) method, fractional-order TV method and Hessian-nuclear-norm regularisation method. Specifically, the proposed method can achieve better and more balanced results in terms of peak-signal-to-noise ratio, convergence rate and restoration quality.

New colour morphological operators on hypergraph are proposed to avoid the loss of details caused by fixed structure element effectively. Hypergraph, the most general structure in discrete mathematics, is also a subset of a finite set. Being a structured representation of information, the ordinary image can be transformed into a hypergraph model, which can integrate hypergraph theory with mathematical morphology theory. As hypergraphs have good performance in structuring information, first of all, this study designs a reasonable algorithm for turning colour images into hypergraph space. Then based on hypergraph theory and colour distance, new colour morphological operators on hypergraph are defined. Experiments show that using the new operator can avoid the loss of detail information and the destruction of colour topology, which improve the precision of image processing.

The fusion of multimodal medical information is considered as an assisted approach for the medical professionals. Computed tomography and magnetic resonance (CT–MR) medical image fusion are able to help the radiologist in precise diagnosis of disease and deciding the required treatment in accord with the patient's condition. Therefore, a cascaded framework is proposed in this study that presents a fusion approach for multimodal medical information in ripplet transform (RT) and non-subsampled shearlet (NSST) domain. The RT and NSST having different features are utilised in a cascade manner that provides several directional decomposition coefficients and increases shift invariance information in the fused images. At the first stage decomposition, a biologically inspired neural model, motivated by novel sum-modified Laplacian and spatial frequency is utilised to fuse the low- and high-frequency coefficients, respectively, and the max fusion rule based on regional energy is utilised at stage 2. This model is used to preserve the redundant information also. The fusion performance is also validated by extensive simulations performed on different CT–MR image datasets of different diseases. Experimental results demonstrate that the proposed method provides better fused images in terms of visual quality along with the quantitative indices compared with several existing fusion approaches.

The impact of radiation dose is directly related to the quality of computed tomography (CT) images. Low-dose CT images are degraded with the noise and other factors. Noise reduction methods are very helpful to enhance the noisy CT images with a possibility to increase the signal-to-noise ratio (SNR) and have a scope for further reduction of radiation dose. In this study, a denoising scheme is proposed which is applicable only for two identical images with uncorrelated noise. In the proposed scheme, a non-local means (NLM) filter is used to denoise the first input image and a wavelet packet thresholding to denoise the second input image. Results of NLM filter are analysed and found excellent for noise suppression but missing the small structures of the input image. To recover that, the proposed scheme is using correlation-based wavelet packet thresholding. The final outcomes of the proposed scheme are excellent in terms of noise suppression and structure preservation. The proposed scheme is compared with existing methods and it is observed that performance of the proposed method is superior to existing methods in terms of visual quality, image quality index, peak SNR and entropy difference.

Hybrid methods filtering of images generally put in collaboration two filters: the first is used for the pre-smoothing and the second to finalise the filtering operation. This approach permits to smoothen the image efficiently while preserving at best its contours and fine structures. Unfortunately, the pre-smoothing generally induces a blur which is not taken into account during the smoothing. In this study, the authors propose a hybrid diffusion method based on the Jeffreys-type equation (JTE). The method proposed provides two steps: a pre-smoothing stage called adaptive-linear diffusion (A-LD) and a smoothing stage called Jeffreys anisotropic diffusion (JAD). A stopping criterion for A-LD, called variance absolute error (VAE) which estimates the noise level, is proposed in order to guarantee the optimality of the smoothing filter. A diffusion function is proposed as well in order to take into account the blurring induced by the pre-smoothing. The proposed hybrid method is compared with the Perona and Malick, telegraph equation, artefact suppressed large-scale non-local means and Xu-AD models. Experimental results show that the proposed hybrid method is efficient both on noise reduction and preservation of contours and fine structures.

Pulse-coupled neural network (PCNN) is a biologically-inspired algorithm suited for image processing. However, determining a set of parameters involved in the alteration of the neural behaviour remains a prevalent research for further application. To apply the model into image segmentation, this study proposes an extended PCNN model by using a strategy of the decision tree, and establishes links between the parameters and image characteristics. Particularly, the adjustable threshold term, interacted with the estimation of the global neural threshold, enables the proposed model to obtain the better results with the use of the fuzzy set theory. Through iterative computation, the proposed model can be considered as a region growing approach for multilevel image segmentation, thus named as an extended PCNN model. Finally, experiments on synthetic and natural images demonstrate the efficiency of the proposed model. Moreover, comparisons with some existing PCNN-based models, and recently graph-based methods, normalised cuts, show that the proposed model can extract regions with more similarity.

This study presents an efficient fast bit rate transcoding method for high-efficiency video coding (HEVC) screen content coding (SCC). It utilises the correlation between the current coding unit (CU) and the corresponding decoded CU to skip the normal intra mode or intra block copy mode. Furthermore, quantisation parameter values are also incorporated into the algorithm for better performance. In addition, a fast scheme is designed to skip partial prediction unit modes to accelerate the entire transcoding process. Experimental results show that, compared with the original HEVC SCC reference software (SCM-3.0), the proposed fast intra transcoding algorithms achieve 45.09% transcoding time reduction with only 1.78% Bjotegaard delta bit rate increment for the all-intra case.

This study is devoted to extracting significant texture features from dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) of the breast using curvelet features and to classify breast masses into malignant and benign using the calculated features. The authors utilised the first generation of curvelet transform in the interpretation of breast tumours on DCE-MRI. The analysis is performed after injecting 23 patients with a contrast agent and 23 mass lesions were extracted from these patients. Then, 288 statistical parameters were extracted by calculating the mean and variance of the curvelet coefficients of tumour texture in sub-band images. Due to a large number of extracted features and the presence of redundant and inter-correlated descriptors, they used a combination of genetic algorithm (GA) and Pearson's correlation for feature selection and a three-layer artificial neural network (ANN) for classification of malignant and benign breast lesions. The GA-ANN model has yielded a good diagnostic accuracy (96%), sensitivity (92%) and specificity (100%). Also, the area under the receiver operating characteristic curve was 0.955. The curvelet transform was able to effectively quantify the distribution of contrast agent in tumour texture, which is different in malignant and benign tumours.

Here, a novel, yet simple, transform-based watermarking scheme for copyright protection has been proposed. The proposed method takes advantage of the rotation invariant property of the ripplet-II transform which makes the proposed method robust against rotation attacks. In the proposed method, firstly a cost function of perceptual transparency is formed in the ripplet-II domain. Secondly, the ridge regularisation constraint is added to the cost function to avoid the singularity problem in the model. In order to obtain the embedding weights, this function is minimised. Therefore, the embedding weights are adaptive to both the host image and watermark. Simulation results demonstrate that the proposed algorithm not only provides an excellent watermark invisibility, but it also shows superior robustness against common image processing operations and acceptable robustness against geometric distortions.

The popular histogram equalisation (HE) technique, which was developed to improve the image contrast, sometimes may also be misused to hide intensity variations in tampered images with ill intention. The authors investigate how existing image forensic techniques may fail to detect HE operation in highly compressed and low-resolution images. They then propose an algorithm to detect whether a given image (either uncompressed or JPEG compressed) has undergone the HE process or not. It is based on the frequency domain analysis of image histogram and exploits the difference in DC and AC coefficients in histogram's discrete Fourier transform. It can detect HE operation even if the image is saved in JPEG format after the equalisation, where most the existing algorithms fail. The extensive computer simulations over large dataset show the effectiveness of the proposed algorithm.

A cubic trigonometric B-spline representation with two parameters is constructed in this work. A soft computing technique, genetic algorithm, is used to find the optimal value of the parameters in the description of B-spline so that the sum square error is minimised. The newly constructed B-spline is then utilised to interpolate two-dimensional digital images. The image quality metrics peak signal-to-noise ratio, structure SIMilarity index, multi-scale structure SIMilarity index and feature SIMilarity index are used to investigate the quality of interpolated digital images. Comparison with already existing image interpolation schemes leads to the conclusion that the proposed image interpolation technique is found to be a valuable scheme for the problems related to digital image interpolation.

The second generation bandelet transform uses the two-dimensional (2D) separable wavelet transform to improve its image denoising and compression performance. However, the 2D separable wavelet transform is not a shift-invariant transform and therefore cannot capture geometric information well. The authors propose a hybrid image denoising method in which the 2D separable wavelet transform in the second generation bandelet transform is replaced with the non-subsampled contourlet transform. The results of the application of the proposed method to several greyscale and colour benchmark images contaminated with various levels of Gaussian white noise and Poisson noise indicate that the proposed method has good peak signal-to-noise ratio and visual quality performance.

This study presents an adaptive image segmentation method based on the fuzzy c-means with spatial information (FCM_S). First, the advantages of the local FCM_S over the global FCM_S and its segmentation characteristics are introduced, on the basis of which, the accumulated local FCM_S is proposed to classify each pixel in an image by using information from different local windows that contain them. The local window size is calculated automatically, and the classification results of all pixels are stored together in the accumulated result. The grey levels of the background and the object pixels in the accumulated image, which is converted from the accumulated result, are distributed around 0 and the maximal grey level. Thus, it can be segmented by the grey level where the change rate of the count of object pixels reaches the minimum. Experiments are performed on 16 images from the Weizmann's database, as well as two real-world and four synthetic images. The results validated that the proposed method can segment images with inhomogeneity well and can gain better area overlap measure when compared with some new segmentation methods. Moreover, the proposed method is parameterless.

Printed circuit board wire segmentation based on computed tomography (CT) image can help subsequently locate and estimate inner faults of circuit in an automatic and non-destructive manner. However, CT imaging is prone to suffer from scattered artefacts, metal artefacts and other interference, destroying compact boundary structures of wires. Wires have the characteristic of dense local distribution, and massive vias, pads, and coppers can appear close to wires, resulting in mazy recognition surroundings. The above-mentioned problems bring great difficulty for high-accuracy recognition and location of wire segmentation. In this study, considering that deep convolutional neural network (DCNN) with powerful feature representation can recognise wires in confused surroundings, and graph cut (GC) model relying on grayscale and local texture information specialises in protecting edge structures of wires, the authors propose an effective framework called DCNN-GC that employs DCNN to obtain global semantic prior to guide the GC model to accomplish satisfactory wire segmentation. The authors qualitative and quantitative results demonstrate outstanding performance, and achieve overwhelming intersection over union compared with traditional and DCNN-based methods.

A hyperspectral anomaly detector, which uses the benefits of a proposed background feature extraction method and the kernel trick, is introduced in this study. The proposed method is differential image based detector (DID). The DID method uses the differential image to estimate the variations of background in the feature extraction phase. The proposed feature extraction method slows the variations of background. So, it suppresses the background signal and highlights the anomalous signal. The separation between anomalous targets and background clutter is increased in the projected feature space. Before feeding the transformed data into the Reed–Xiaoli (RX) detector, the kernel learning method is applied. The kernel technique transforms the projected data with a linear non-Gaussian model to a potentially high-dimensional feature space with the non-linear Gaussian domain. Experiments are conducted on two real hyperspectral images. The experimental results indicate that the proposed DID method outperforms RX and some state-of-the-art anomaly detection approaches.

Perceptual quality assessment of stereoscopic images is a challenge in three-dimensional video systems. Existing studies suggest that simply averaging the quality of left and right views can effectively predict the quality of symmetrically distorted stereoscopic images, but prediction deviation occurs in the case of asymmetrically distorted stereoscopic images. Most previous stereoscopic image quality assessment (SIQA) methods have been based only on the luminance component of the images; in addition, the basis of human visual perception is critical to image quality assessment and lies on the low-dimensional manifold. Inspired by this, a new perceptual SIQA method is proposed, which includes two stages: training stage and quality prediction stage. In the training stage, the authors apply Tucker decomposition to RGB images to reduce dimensions along colour channels to produce training sets, and the projection matrix is obtained through manifold learning. In the quality prediction stage, considering the binocular visual characteristics of visual perception, the overall stereoscopic estimate depends on the monocular image quality via a local energy ratio based pooling strategy and cyclopean based binocular quality. Extensive experiments on three available benchmark databases demonstrate that the proposed metric has better performance and achieves highly consistent alignment with subjective assessment compared with state-of-the-art SIQA metrics.

Convolutional neural network (CNN) has proven to be a highly efficient approach to face recognition. In this study, the authors introduce a new layer to embed the patch strategy in convolutional architectures to improve the effectiveness of face representation. Meanwhile, a multi-branch CNN is constructed to learn features of each cropped patch by the patch strategy and then fuses all the patch features together to form the entire face representation. Compared with the traditional patch methods, their approach has the advantage that no extra space is needed to store the facial patches since the images are cropped online. Moreover, due to the end-to-end training, this approach makes a better use of the interactions between global and local features in the model. Two baseline CNNs (i.e. AlexNet and ResNet) are used to analyse the effectiveness of their method. Experiments show that the proposed system achieves comparable performance with other state-of-the-art methods on the labelled faces in the wild and YouTube face verification tasks. To ensure the reproducibility, the publicly available training set CASIA-WebFace is used.

Face recognition technique has obtained great progress and excellent results on public data sets. However, traditional algorithms suffer from various changes such as illumination, expression, and misalignment in practical applications. To solve these problems, this study proposes a novel face recognition algorithm simultaneously resolves these challenges. The key idea is reducing the influence of illumination and expression through the aligning procedure. As a result, illumination, expression, and misalignment can be greatly ignored in the recognition procedure. The contributions of this study are two folds. (i) The construction of the shape constrained illumination pattern (SCIP), which models the illumination variation with robustness to expression change. (ii) SCIP-based face recognition algorithm which can deal with illumination, expression, and image misalignment simultaneously. Systematic evaluations conducted on public databases demonstrate that the proposed algorithm is robust to illumination, expression, and misalignment with better performance than state-of-the-art algorithms.