Abstract: According to the development of the real fractional differential and its applications in the modern signal processing, the authors extend it to quaternion body and put forward a new concept: quaternion fractional differential (QFD), and apply it to edge detection of colour image. This method is called edge detection based on QFD. Simulation experiments indicate that this method has special advantages. Furthermore, the authors give an indicator to evaluate the effectiveness of different edge filters. Comparing with Sobel and mix edges of real fractional differential to every channels of colour image, they discover that QFD has fewer false negatives in the textured regions and is also better at detecting edges that are partially defined by texture, which means the authors can obtain much better results in the interesting regions using QFD and is more consistent with the characteristics of human visual system.

Abstract: Segmenting iris texture from an input image is an important step for recognising iris pattern. It is still a difficult task to localise available texture regions from non-ideal iris images captured in non-cooperative situations such as lighting variations, on-the-move and off-angle view. To address this problem, this study presents a novel algorithm for accurate and fast iris segmentation. An adaptive mean shift procedure is built to find the rough position of the iris centre. According to the localisation result, a circle is set as the initial iris contour. After combining the statistical texture prior modelled as Markov random field, a merged active contour model is established in terms of level set theory. Under the MAC model, the initial contour is iteratively driven to real iris boundaries. During the curve evolving process, eyelids, eyelashes, reflections and shadows can be simultaneously detected and labelled in iris regions. The novelty of the proposed method mainly includes developing a new modified mean shift procedure for fast and robust iris localisation, and successfully incorporating the local probabilistic prior, boundary and region information into the designed active contour model for accurate texture segmentation. Extensive experimental results on various challenging iris images show that our method can effectively and accurately perform iris segmentation with low computational complexity and has very promising applications in non-cooperative recognition systems.

Abstract: Visual and infrared cameras have complementary properties and using them together may increase the performance of object detection applications. Although the fusion of visual and infrared information results in a better recall rate than using only one of those domains, there is always a decrease in the precision rate whereas the infrared domain on its own always has higher precision. Thus, the fusion of these domains is meaningful only for a better recall rate, which means that more foreground pixels are detected correctly. This study presents a new computationally more efficient and simpler method for extracting the complementary information from both domains and fusing them to obtain better recall rates than those previously achieved. The method has been tested using a well-known database and a database created for the study and compared with earlier fusion methods.

Abstract: Traditional watermarking schemes are sensitive to geometric distortions, in which synchronisation for recovering embedded information is a challenging task because of the disorder caused by rotation, scaling or translation (RST). The existing RST-resistant watermarking methods still have limitations with respect to robustness, capacity or fidelity. In this study, the authors address several major problems in RST-invariant watermarking. The first point is how to take advantage of the high RST resilience of scale-invariant feature transform (SIFT) features, which show good performance in terms of RST-resistant pattern recognition. Since many keypoint-based watermarking methods do not discuss cropping attacks, the second issue discussed in this study is how to resist cropping using a human visual system (HVS), which also helps us to eliminate computational complexity. The third issue is the investigation of an HVS-based watermarking strategy for extracting only feature points in the human attentive area. Lastly, a variable-length watermark synchronisation algorithm using dynamic programming is proposed. Experimental results show that the proposed algorithms are practical and show superior performance in comparison with many existing works in terms of watermark capacity, watermark transparency, and the resistance to RST attacks.

Abstract: Visual sensing is an attractive approach to detect the weld position in an arc-welding process, which provides information for seam tracking However, it is difficult to accurately detect the weld position adjacent to a molten pool because of strong arc disturbances A novel algorithm based on the weld-pool image centroid is presented to improve the seam-tracking ability The molten pool images are taken by a camera arranged ahead of the welding torch and the centroid is extracted as a parameter to detect the weld position It is worth noting that the centroid corresponds to the thermal distribution of the molten pool affected by the offset between the arc and the seam centreline Therefore the offset between the arc and the seam centreline can be estimated by this centroid The least square linear regression method is employed to correlate the relationship between the centroid and the offset under different welding conditions For further analysis of the centroid characteristics, a non-linear neural network is designed with three input variables which are the position, displacement and moving velocity of the centroid, respectively This neural network model shows higher accuracy of weld detection In comparison with directly detecting the weld position, the centroid can be measured and computed easily This algorithm is expected to provide a promising vision model to improve the accuracy of seam tracking in real time, and subsequently to ensure good welding quality

Abstract: The authors propose a novel image segmentation technique based on the non-parametric clustering procedure in the discretised colour space. The discrete probability density function is estimated in two steps. Multidimensional colour histogram is created, which is afterwards used to acquire final density estimate using the variable kernel density estimation technique. Segmentation is obtained by mapping revealed range domain clusters to the spatial image domain. The proposed method is highly efficient, running in time linear to the number of the image pixels with low constant factors. The output of the algorithm can be accommodated for a particular application to simplify the integration with other image processing techniques. Quantitative evaluation on a standard test dataset proves that the quality of the segmentations provided by the proposed method is comparable to the quality of the segmentations generated by other widely adopted low-level segmentation techniques, while running times are several times faster.

Abstract: A novel, efficient and simple impulse noise detector for switching median filter is proposed in this study. The proposed method compares the difference value of the current pixel with the brightest and the darkest pixels in its working window and uses the difference value to determine whether the current pixel is corrupted by impulse noise. To avoid the false alarm generated in this first stage, the candidate pixels are passed through a second stage using local statistics. This new technique can remove the impulsive noise from corrupted images (grey scale and colour) efficiently and requires no previous training and noise statistics or strength for that purpose. For performance evaluation, four impulse noise models (Ng and Ma, 2006) are considered and noise models 3 and 4 are generalised to expand its application. Proposed algorithm is blind to the noise model and the amount of noise introduced. Quantitative and qualitative analysis performed on standard images, show that proposed method can detect impulse noise very efficiently under a wide range (up to 90%) of noise density. The proposed method performs well in terms of low miss and false detection and high peak-signal-to-noise ratio (PSNR) value. In comparison to all examined algorithms, the proposed technique performs favourably and outperforms the competitors for all the noise models. The gain is most prominent in case of impulse noise having spread as in case of noise models 3 and 4 (Ng and Ma, 2006).

Abstract: Mode mixing, boundary effects, necessary extrema lacking and so on are the main problems involved in bi-dimensional empirical mode decomposition (BEMD). The study presents an improved BEMD based on 2D-assisted signals: 2D Gaussian noises. Firstly, the given 2D Gaussian noise and its negative counterpart are added to the original image, respectively, to construct the two images to be decomposed. Secondly, the decomposed intrinsic mode functions (IMFs) from the two images are added together to obtain the IMFs, in which the added noises are cancelled out with less mode mixing and boundary effects. The other contribution of the method lies in its overcoming of the problem of necessary extrema lacking that the previous BEMD fails. Some instructive conclusions are obtained in the improved BEMD. Lastly, the efficiency and performance of the method are given through theoretical analysis and its application in image enhancement, which outperforms some previous approaches.

Abstract: This study introduces a block-based motion estimation algorithm based on projection with adaptive window-size selection. Using one-dimensional (1D) projection, the original 2D block-matching problem is translated to a simpler 1D matching problem which eliminates the majority of potential pixel participation. The authors propose an efficient fast-motion estimation method using an intensity gradient algorithm to reduce the encoding time. The proposed algorithm estimates the texture direction of the block, and then a horizontal or vertical projection is employed for the sum of absolute difference calculation. Experimental results show that the proposed algorithm achieves a greater than 60% time reduction with negligible peak signal to noise ratio (PSNR) loss and BR increments.

Abstract: In motion estimation, the sub-pixel matching technique involves the search of sub-sample positions as well as integer-sample positions between the image pairs, choosing the one that gives the best match. Based on this idea, this work proposes an estimation algorithm, which performs a 2-D correspondence search using a hierarchical search pattern. The intermediate results are refined by 3-D cellular automata (CA). The disparity value is then defined using the distance of the matching position. Therefore the proposed algorithm can process uncalibrated and non-rectified stereo image pairs, maintaining the computational load within reasonable levels. Additionally, a hardware architecture of the algorithm is deployed. Its performance has been evaluated on both synthetic and real self-captured image sets. Its attributes, make the proposed method suitable for autonomous outdoor robotic applications.

Abstract: A new approach has been proposed to improve the computational performance of denoising in which adaptively defined learning step size has been used for tuning the parameter of the thresholding function of wavelet transform-based thresholding neural network (WT-TNN) methodology. In this approach, steepest gradient-based learning step size of WT-TNN methodology are changed to the proposed adaptively defined learning step size for tuning the parameters of thresholding function. The results of the image enhanced by such adaptive learning step size exhibit the increase in the speed of learning and improved edge preservation feature. Further, the learning time has also become independent of noise level and initial values of learning parameters.

Abstract: This study presents an efficient saliency model mainly aiming at content-based applications such as salient object segmentation. The input colour image is first pre-segmented into a set of regions using the mean shift algorithm. A set of Gaussian models are estimated on the basis of segmented regions, and then for each pixel, a set of normalised colour likelihood measures to different Gaussian models are calculated. The colour saliency measure and spatial saliency measure of each Gaussian model are evaluated based on its colour distinctiveness and the spatial distribution, respectively. Finally, the pixel-wise colour saliency map and spatial saliency map are generated by summing the colour and spatial saliency measures of Gaussian models weighted by the normalised colour likelihood measures, and they are further combined to obtain the final saliency map. Experimental results on a dataset with 1000 images and ground truths demonstrate the better saliency detection performance of our saliency model.

Abstract: More people are studying digital video stream transference via networks. However, frequent Internet use increases the requirement for copyright protection and security. As a consequence, to prevent video streams that belong to rightful owners from being intentionally or unknowingly used by others, information protection is indispensable. The authors propose a novel method for video watermarking that is specifically designed for H.264 video. For the experiment, a low-energy signal can relatively guard against low-pass filter attacks. Conversely, a high-energy signal in the host signal can relatively guard against the high-frequency noise attack. In view of these facts, the proposed system design embedding algorithm provides high-energy and low-energy blocks. The blocks in the host image frame are divided into two different groups by estimating the block energy. The existing singular value decomposition methods were employed to calculate the watermark information. In order to enhance the security, the proposed system also employs torus automorphisms to encrypt the watermark. To achieve better robustness, the encrypted results use secret image sharing technology embedded into different I-frames in the video stream.

Abstract: Noise gradient is reduced while image details gradient is also reduced by a filter. For the image corrupted by impulse noise, a novel approach of noise gradient reduction is proposed based on a pair of morphological dual operators. The noise image is filtered by a pair of morphological dual operators respectively, and then the two filtered images are provided with the complementary characteristics of the noise gradient position. This feature results from the unsymmetric behaviour of the pair of morphological dual operators, and it can be applied to reduce the noise gradient effectively. This approach is presented in detail and the experimental results show that the approach not only reduces noise gradient effectively, but also maintains image details gradient. Compared with the classical morphological dual operators, the generalised morphology dual operators have smaller root mean square error on the premise of the close computation and time.

Abstract: Recent researches have shown that the adaptive directional lifting (ADL) can represent edges and textures in images effectively. This makes it possible to separate noise from image signal distinctly in image denoising. However, a key issue named orientation estimation for ADL becomes inefficient and error prone in the noised circumstance. The authors propose a robust adaptive directional lifting-based (RADL) wavelet transform for image denoising by constructing ADL in an anti-noise way. In our method, a simple model of pixel pattern classification is incorporated into orientation estimation module to strengthen the robustness of this algorithm. Moreover, instead of determining the transform strategy based on sub-blocks, RADL is performed on pixel-level to pursue better denoising results. Experimental results show that the proposed technique demonstrates both PSNR and visual quality improvement on images with rich textures.

Abstract: Combined discrete wavelet transform-fan beam transform (DWT-FBT) has been explored as a new possible domain for colour image watermarking. The two schemes proposed in the combined domain are (i) wavelet fan beam watermarking on luminance and chrominance and (ii) wavelet fan beam watermarking on chrominance alone. After the application of DWT on the host image and after careful selection of the suitable band of wavelet coefficients for applying FBT, watermarking is done by altering the fan beam transformed coefficients. The schemes proposed achieve a high data embedding capacity in addition to robustness to attacks and produce watermarked images of high quality. Results of the proposed schemes are compared with the two existing DWT-discrete cosine transform watermarking schemes to show its effectiveness.

Abstract: This study presents a robust and reliable method of human posture recognition for visual surveillance systems. In order to recognise the human body, a recognition method is developed based on the skeleton of moving object. To obtain the skeleton of object, the authors describe some thinning algorithms for binary images, including one pass thinning algorithm, Zhang's thinning algorithm, Rosenfeld's thinning algorithm and a new thinning algorithm. Three performance measurements are chosen to evaluate these thinning algorithms. Comparing the performance results the authors found that the proposed thinning algorithm had managed to produce several improvements, including high thinness, connectivity, robustness to noise and low time consuming. Moreover, the skeleton obtained by the proposed thinning algorithm is one-pixel width and more smooth. Next, three different postures such as standing, bending and crawling will be estimated by using support vector machines as a classifier, which the histograms of horizontal and vertical projections are selected to define the feature. Finally, experimental results demonstrate that the human body and posture estimation algorithm have a robust and real-time performance, and is useful for the discrimination of human postures.

Abstract: Many video object tracking systems use block matching algorithm (BMA) because of its simple computational structure and robust performance. The BMA, however, exhibits fundamental limitations resulting from non-rigid shapes and similar patterns to the background. The authors propose a combined shape and feature-based non-rigid object tracking algorithm, which is tightly coupled with an adaptive background generation to overcome the limit of block matching. The proposed algorithm is robust to the object's sudden movement or the change of features. This becomes possible by tracking both feature points and their neighbouring regions. Combination of background and shape boundary information significantly improves the tracking performance because the target object and the corresponding feature points on the boundary can be easily found. The shape control points (SCPs) are regularly distributed on the contour of the object, and the authors compare and update the centroid during the tracking process, where straying SCPs are removed, and the tracking continues with only qualified SCPs. As a result, the proposed method becomes free from potential failing factors such as spatio-temporal similarity between object and background, object deformation and occlusion, to name a few. Experiments have been performed using several in-house video sequences including various objects such as a moving robot, swimming fish and walking people. In order to demonstrate the performance of the proposed tracking algorithm, a number of experiments have been performed under noisy and low-contrast environment. For more objective comparison, performance evaluation of tracking surveillance 2002 data sets were also used.

Abstract: Different substrate types, such as seagrass, sand, stone and silt, can be identified by detecting substrate reflectivity. In this study, high-resolution remote sensing data of Quickbird were used to map seagrass distribution in the northeast coast of the Xincun Bay. Radiance transfer model in optical shallow water was used to retrieve bottom reflectivity; relationship between hyperspectral band and leaf area index (LAI) was used for processing satellite remote sensing data to retrieve distribution and density of submerged seagrass. Hyperspectral results showed that spectral bands at 550, 650 and 675 nm were sensitive to LAI. Based on the relationship between LAI and hyperspectral bands, Bands 2 and 3 of Quickbird could be used for retrieving seagrass distribution. Satellite results showed that seagrass was mainly distributed in clusters in the northeast coast of the Xincun Bay, tens of metres away from the bank. Density gradient was very clear, with coverage between 40 and 60 occupied by a great part of seagrass bed and with coverage greater than 80 distributed in stripe pattern in the main seagrass bed. The accuracy of seagrass retrieval was more than 80 , and the density of seagrass could also be distinguished evidently.

Abstract: A new relevance feedback (RF) approach for content-based image retrieval (CBIR) is presented, which uses Gaussian mixture (GM) models as image representations. The GM of each image is obtained as an adaptation of a universal GM which models the probability distribution of the features of the image database. In each RF round, the positive and negative examples provided by the user until the current round are used to train a support vector machine (SVM) to distinguish between the relevant and irrelevant images according to the preferences of the user. In order to quantify the similarity between two images represented as GMs, Kullback–Leibler (KL) approximations are employed, the computation of which can be further accelerated taking advantage from the fact that the GMs of the images are all refined from a common model. An appropriate kernel function, based on this distance between GMs, is used to make possible the incorporation of GMs in the SVM framework. Finally, comparative numerical experiments that demonstrate the merits of the proposed RF methodology and the advantages of using GMs for image modelling are provided.

Abstract: In order to implement 3D measurement in different measurement systems with line structured-light sensor, the extrinsic matrix for transforming the position between the structured-light sensor and the measurement platform must be calibrated. An extrinsic calibration method, achieved by measuring the fixed point (the centre of a standard ball) of the measuring system with the sensor in different positions, was proposed. In order to calculate the coordinate of the ball centre in the laser plane frame more accurately, the images containing fewer pixels were filtered and the noise around the slice was removed. Experimental study was carried out on a four-axis coordinate measuring machine (CMM) with a ball supplied by CIMCORE and showed that the method is simple; it can also be used in a non-orthogonal measurement system.

Abstract: This study presents a precise, one-pass block classification algorithm for efficient coding of computer screen images like power-point presentations, webpages and wall papers. The objective is to minimise the loss of visual quality of text during compression by separating text information which needs high spatial resolution than the pictures and background. It segments computer screen images into text/graphic, picture/background blocks by computing the statistical feature based on discrete wavelet transforms coefficients in the detail sub-bands of each 8×8 block, and then compresses the text/graphics pixels losslessly with a two-mode block prediction coding and the background pixels with the lossy JPEG algorithm. The proposed scheme performs accurate block classification of text information with different fonts, sizes and ways of arrangement from the background image, so that text/graphics blocks are compressed at higher quality than background image blocks. Experimental results show that the proposed method minimises block classification error and improves the value of peak signal-to-noise ratio significantly than standard JPEG, JPEG-2000 and H.264/AVC-I, while keeping competitive compression ratio and visually lossless quality of text information.

Abstract: In this study, an illumination-tolerant face recognition algorithm is proposed. This work highlights the significance of matrix polar decomposition for illumination-invariant face recognition. The proposed algorithm has two stages. In the first stage, the authors reduce the effect of illumination changes by weakening the discrete cosine transform coefficients of block intensities using a new designed quantisation table. In the second stage, the unitary factor of polar decomposition of the reconstructed image is used as a feature matrix. In the recognition phase, a novel indirect method for measuring the similarities in feature matrices is proposed. The nearest-neighbour rule is applied to the matching. The authors have performed some experiments on several databases to evaluate the proposed method in its different aspects. Experimental results on recognition demonstrate that this approach provides a suitable representation for illumination invariant face recognition.

Abstract: An efficient intra-mode search algorithm is proposed to reduce the computational complexity of inter-frames for the H.264/AVC video encoding system. Each intra-mode search procedure usually causes a high degree of complexity in an inter-frame because of the inter-frame prediction modes that operate with the inter-mode decision procedure. To decrease this computational burden, the authors propose an adaptive thresholding algorithm based on distribution characteristics of the sum of the absolute differences (SAD) of the best inter-mode when the intra-mode is the final coding mode. The authors also include a simple refinement process by using the spatial correlation between neighbouring macro-blocks (MBs) and the current MB. The performance of the proposed algorithm is verified through comparative analysis of encoding time, loss of quality and the bit increment.

Abstract: In this study, the authors propose a novel algorithm for removing complex shadows from textured surfaces within a single image. Previous shadow removal methods have attempted to solve this problem under a range of criteria, such as the type and number of light sources and shadow structure. Some methods require that the shadow umbra and penumbra be well defined. They have been shown to work best when the umbra is relatively large and when the penumbra is of a limited size and shape. However, these methods can fail if the shadow shape or form diverges from the norm. Many methods also require user intervention to locate whole or parts of the shadow. The authors propose a flexible shadow removal model which is capable of functioning without these limiting assumptions, or the need for user intervention. It is also capable of dealing with shadows cast on a wide range of textured surfaces. It uses a directional differential filter along with directional smoothing to find the shadow and a thin plate reconstruction model to remove shadows from an image surface. Results have shown that the proposed algorithm generates high-quality shadow-free images over a range of scenarios, such as multiple light sources, occlusions and textures.

Abstract: Similarity measurement based on mutual information maximisation has been successful applied in image registration. However, it costs a lot of computation time and the interference of local maxima in the search process always makes the registration search into local maxima that may cause misregistration. In order to eliminate these shortcomings, a novel image registration method is presented in this study. In the method wavelet multi-scale product is calculated to extract the feature points and angle information of the two input images, then a new criterion of registration – criterion of feature point pair mutual information – is defined to acquire corresponding matching points. In the experiments, our method and image registration methods based on correlation criteria and alignment metric criteria are tested for comparison. Experimental results show that the registration of our method outperforms that of the other two methods and the registration errors are obviously reduced. The errors of coordinates are lower compared to the errors of 78% of the other two methods and the errors of rotation angle are lower compared to the errors of 6% of the other two methods. The seams of the registration image results are very smooth and the transition zones are very uniform.

Abstract: A novel image enlargement method based on a modified interpolatory subdivision scheme is proposed in this study. The subdivision scheme is a modification from the 4-point interpolatory subdivision by substituting the interpolation rule for a tangent-constrained Hermite interpolation and in surface case the subdivision is derived from a Ferguson patch. By estimating the gradients of the Ferguson patch, the authors present an image enlargement algorithm preserving the sharp edges. Benefit from the advantages of subdivision, this algorithm runs fast. Numerical experiments illustrate the efficiency of the novel method.

Abstract: Millimetre-wave (MMW) radiometric imaging has established itself in a wide range of military and civil applications, but the fact that the radiometric temperature contrast between different objects indoors is low restricts the commercial application of MMW radiometric imaging to some extent. Based on the particular analysis of MMW radiometric imaging model suitable for concealed contraband detection on personnel, the method of adopting MMW noise source to illuminate the scene of being imaged to improve the radiometric temperature contrast, just like the flash lamp of camera, is proposed in this study. In addition, the choice of both the ways of illumination and the excess noise ratio (ENR) of the noise source is analysed. Then the image degradation model is discussed and the projection onto the convex sets (POCS) algorithm for MMW radiometric image super-resolution restoration is depicted in detail. In addition, a series of experiments are carried out both indoors and outdoors, with the dual-channel radiometric imaging system that has been implemented at W-band with the mode of two-dimensional planar scanning. The experimental results are presented and analysed to validate the significance of the imaging model and it is indicated that the radiometric imaging system is capable of detecting metal objects hidden underneath clothing, and the method of adopting noise illumination can improve the quality of radiometric images.

Abstract: The thick-wall ECT system applied in high pressure is introduced, and the pipe wall thickness of ECT sensor is 5 mm. To analyse the effect of pipe wall thickness on ECT sensor, radius electrode ratio is defined as the ratio of inner and outer radius of pipeline. ANSYS and Matlab are combined to simulate sensor characteristic. Linear back projection algorithm is adopted to reconstruct the images of stratified flow, core flow and annular flow with different radius electrode ratios. The results of simulation indicate that there is a small image error under high radius electrode ratio, and the images reconstructed from stratified flow are better than those from core flow and annular flow. The independently developed ECT system is applied to identify the flow regime of gas solid flows in horizontal pipeline with 10 mm inner diameter in a pulverised coal dense-phase pneumatic conveying experimental setup. The pressure is up to 4.0 MPa and the solid gas ratio is up to 11.73 kg/kg. The experimental results of eight-electrode ECT with thick pipeline sensor, which is used on the pulverised coal dense-phase pneumatic conveying experimental setup, are given. Finally, the Fluent simulation of concentration distribution at the similar condition is introduced to validate the ECT imaging result.