The deployment of autonomous robots in human environments requires an understanding of social interactions and the factors that influence them. Human-robot proxemics is an important factor that impacts interactions, and modeling personalized proxemic behavior has always been a challenge, as it depends on multiple user attributes, including gender, age, and height. In this paper, we propose a novel approach that uses rubber-sheet transformation models to represent human-robot proxemics. We do this by collecting human-robot interpersonal distance data from 20 users and model it with respect to their height, age, gender, and the angle at which the robot approaches. We present an evaluation of the model, and the outcome of our results, which show a promising approximation of proxemic distances based on different user attributes. Finally, we provide a coefficient table for rubber-sheet models to lay the foundation for personalized human-robot proxemics and outline future research directions.

This paper proposes an effective and robust texture feature descriptor to classify mammographic images into different breast density categories. Accurate breast density based categorization of images plays an important role in the risk assessment at early stages of breast cancer. Based on the commonly used local binary patterns (LBP), we investigate its variant method local quinary patterns (LQP) for considering more details of texture features. The rotation invariant approach with different concerned numbers of spatial transitions is used to extend LQP to rotation invariant LQP (RILQP). The proposed method recognizes more texture patterns and reduces the high dimensionality of its feature vector significantly, which make it a robust texture descriptor. In addition, in the breast density classification task, this paper also investigates the influence to classifying results by using resized mammogram images. Two mammogram datasets, INBreast and MIAS, are used in our experiments to test the proposed method. Comparing to state-of-the-art methods, competitive classifying results are observed using the RILQP method, with classification accuracies of 82.50% and 80.30% on INBreast and MIAS respectively. Comparative analysis also indicates that the proposed method outperforms other methods statistically.

Detecting individual micro-calcifications (MCs) in mammograms is a challenging problem due to heterogeneous properties and diverse composition of breast tissues. False positives (FPs) are therefore a common occurrence in the outputs of different detectors. This paper focuses on FP reduction and improvement of the final MCs detection accuracy in mammograms. The proposed method uses a combination of a MC detector which outputs a patch set containing candidate MC spots, and alpha images derived from multifractal analysis to enhance texture features of MC spots in each target patch. For further highlighting the texture features, a Weber's law based approach is used to construct a new multifractal measure and the corresponding alpha patches. In order to distinguish MC spots from the candidate set, a convolutional neural network (CNN) classifier is designed to process original mammogram patches and corresponding alpha patches together for classifying suspicious MC spots to true positive group or false positive group. Multifractal features contained in alpha images are fed into the proposed CNN model, which facilitate learning richer representations for MCs in local regions and presenting better classification performance. A digital mammogram dataset, INbreast, is used to test the proposed method. Experimental results are evaluated using free-response receiver operating characteristic (FROC) and area under the FROC curve (AUC). In our experiments, a desirable classification performance is observed after using the new alpha patch set in the designed CNN classifier, and the general MC detection results based on individual mammograms in a test set demonstrate that the proposed method reduces FP numbers and improves the MC detection accuracy effectively.

Substantia nigra (SN) and red nucleus (RN) located in midbrain are integral in the study of brain disease such as Parkinson's disease (PD). The automatic segmentation of SN and RN in high-resolution quantitative susceptibility mapping (QSM) images can aid in PD characterization and progression. However, only a few methods have been proposed to segment them, owing to the recent development of high quality imaging. Therefore, we describe a novel method for the segmentation of SN and RN in QSM images using contrast enhancement, level set method, wavelet transform and watershed transform. The segmentation performance is evaluated in 20 subjects containing both healthy and PD patients. The results of the proposed segmentation method were closer to the manual segmentation performed by the radiologist than the popular level set methods. The Dice coefficient of the left SN and right SN were 0.77 ± 0.09 and 0.78 ± 0.07 respectively while the Dice for the left RN and right RN were 0.80 ± 0.08 and 0.77 ± 0.08 respectively.

Segmentation of smaller brainstem nuclei like substantia nigra (SN) is one of the first stage in computer-aided diagnosis to investigate the Parkinson's disease (PD) characteristics and progression. PD patients generally have smaller SN in comparison to healthy individuals. There are very few automated methods proposed for SN segmentation and most of them require quite a few number of reference images for increasing the accuracy. Therefore, we propose an improved algorithm for the segmentation of SN using level set method and dual-tree complex wavelet transform (DT-CWT). The proposed level set method uses the local image information and maintains the regularity of the level set function for accurate computation, thus precluding the expensive re-initialization process. DT-CWT is used to solve the problem of over-segmentation, also resulting in holes, and to smooth the jagged outputs generated by the level set method. The experimental results suggest that the proposed algorithm is able to segment SN and is closer to manual delineation.

Plankton are very small marine organisms found in large numbers in aquatic environments. Some of the plankton are bioluminescent, which makes them glow. This characteristic of plankton interferes with the vision systems of remotely operated underwater vehicles (ROVs). As ROVs are becoming increasingly used for various underwater exploration and farming activities, there is a need for creating visually realistic simulation of underwater environments, and plankton is one of the key influencing organic material of that environment. In this paper, we deal with two different types of plankton, phytoplankton and zooplankton and their different optical characteristics. We propose novel approaches for modelling zooplankton by using a particle engine, and phytoplankton using exponential height fog. These models enhance the realism of simulated underwater environments which was successfully demonstrated by visual comparison of the simulated images generated using Unreal Engine 4 with real images.

Computer assisted evaluations of Whole Slide Images (WSI) of histopathological slides require robust biomarker-specific feature descriptors for accurate grading and classification. Considering the large amount of processing involved in analysing WSIs, training and classification, it is important to have an optimized set of features that closely represent the characteristics of the biomarkers used by pathologists in manual assessments. In this paper, we consider the problem of classifying WSIs of ImmunoHistoChemistry (IHC) stained slides for automated breast cancer grading. We use a combination of intensity and texture features derived from the input image at different saturation levels, and show its effectiveness in a Neural Network architecture for classifying the image into one of the four HER2 scores. The paper also presents three configurations for the neural network and gives comparative analysis showing the variations of classification accuracy with respect to changes in the configuration and the learning rate.

The paper presents a novel approach for suppressing speckle noise at the same time preserving edge information effectively in ultrasound images for better clinical analysis and problem identification. The framework includes the modified adaptive Wiener filter (MAWF) along with the Canny edge detection method and enhanced steerable pyramid transformation (SPT) algorithm. The Canny algorithm is used to detect the true edges from the noisy ultrasound (US) image, and the MAWF algorithm smoothens the speckle effect without affecting the edge information which is preserved separately and added to the final output. The discrete Fourier transform (DFT) is used to extract the low and high frequency coefficients. Unlike other multiresolution techniques used for speckle suppression, the proposed method uses the steerable pyramid transformation technique based on high frequency components extracted using DFT for image enhancement. The coherence component extraction (CCE) method enhances the overall texture and edge features of the image even in the darker portions of the image. The output of this stage is finally combined with the stored edge information. This paper also presents experimental results to show that the proposed technique outperforms other state-of-art techniques in terms of peak signal to noise ratio, structural similarity index, and universal quality index.

High Resolution Computed Tomography (HRCT) images of the lungs contain a huge amount of geometrical information that could be effectively extracted to construct a mesh surface of the underlying geometry to a high level of detail. In this paper, we consider this as a problem of constructing 3D surfaces from contours of axial slices. While many reconstruction algorithms adopt volumetric approaches and ray casting methods, we propose a novel algorithm for automatic segmentation of large volume sets, and a contour-based construction of a three-dimensional mesh representation that may be easily used in any rendering application or combined with larger meshes of anatomical parts. Several acceleration structures to reduce the complexity of the algorithm are also presented. Experimental results show that the proposed method provides a high level of detail and quality of rendering that may find useful applications in the field of visualization and graphics.

This study introduces a new set of orthogonal polynomials and moments and the set's application in signal and image processing. This polynomial is derived from two well-known orthogonal polynomials: the Tchebichef and Krawtchouk polynomials. This study attempts to present the following: (i) the mathematical and theoretical frameworks for the definition of this polynomial including the modelling of signals with the various analytical properties it contains, as well as, recurrence relations and transform equations that need to be addressed; and (ii) the results of empirical tests that compare the representational capabilities of this polynomial with those of the more traditional Tchebichef and Krawtchouk polynomials using speech and image signals from different databases. This study attempts to demonstrate that the proposed polynomials can be applied in the field of signal and image processing because of the promising properties of this polynomial especially in its localisation and energy compaction capabilities.

Multifractal analysis has recently found applications in the field of automatic classification and content based retreival of tissue and cell images. The classification accuracy in such methods largely depends on the invariant characteristics of the spectra computed for a chosen intensity measure. It is therefore necessary to carry out a detailed analysis of suitable measures and the invariance of the resulting spectra. This paper discusses some key implementation aspects of multifractal based classification techniques, gives a comparative analysis of different types of intensity measures, and presents experimental results showing the corresponding inter-class and intra-class variations in the multifractal spectra.

Geometric moments have been used in several applications in the field of Computer Vision. Many techniques for fast computation of geometric moments have therefore been proposed in the recent past, but these algorithms mainly rely on properties of the moment integral such as piecewise differentiability and separability. This paper explores an alternative approach to approximating the moment kernel itself in order to get a notable improvement in computational speed. Using Schlick's approximation for the normalized kernel of geometric moments, the computational overhead could be significantly reduced and numerical stability increased. The paper also analyses the properties of the modified moment functions, and shows that the proposed method could be effectively used in all applications where normalized Cartesian moment kernels are used. Several experimental results showing the invariant characteristics of the modified moments are also presented.

The discrete Tchebichef transform (DTT) is a transform method based on discrete orthogonal Tchebichef polynomials, which have applications recently found in image analysis and compression. This letter introduces a new fast 4 x 4 forward DTT algorithm. The new algorithm requires only 32 multiplications and 66 additions, while the best-known method using two properties of the DTT requires 64 multiplications and 96 additions. The proposed method could be used as the base case for recursive computation of transform coefficients. Experimental results showing performance improvement over existing techniques are presented.

Discrete orthogonal moments have several computational advantages over continuous moments. However, when the moment order becomes large, discrete orthogonal moments (such as the Tchebichef moments) tend to exhibit numerical instabilities. This paper introduces the orthonormal version of Tchebichef moments, and analyzes some of their computational aspects. The recursive procedure used for polynomial evaluation can be suitably modified to reduce the accumulation of numerical errors. The proposed set of moments can be used for representing image shape features and for reconstructing an image from its moments with a high degree of accuracy.

Pseudo-Zernike moments have better feature representation capability, and are more robust to image noise than those of the conventional Zernike moments. However, due to the computation complexity of pseudo-Zernike polynomials, pseudo-Zernike moments are yet to be extensively used as feature descriptors as compared to Zernike moments. In this paper, we propose two new algorithms, namely coefficient method and p-recursive method, to accelerate the computation of pseudo-Zernike moments. Coefficient method calculates polynomial coefficients recursively. It eliminates the need of using factorial functions. Individual order or index of pseudo-Zernike moments can be derived independently, which is useful if selected orders or indices of moments are needed as pattern features. p-recursive method uses a combination of lower order polynomials to derive higher order polynomials with the same index q. Fast computation is achieved because it eliminates the requirements of calculating polynomial coefficients, Bpqk, and power of radius, rk, in each polynomial. The performance of the proposed algorithms on moment computation and image reconstruction, as compared to those of the present methods, are experimentally verified using a set of binary and grayscale images.

This paper introduces a new set of orthogonal moment functions based on the discrete Tchebichef polynomials. The Tchebichef moments can be effectively used as pattern features in the analysis of two-dimensional images. The implementation of the moments proposed in this paper does not involve any numerical approximation, since the basis set is orthogonal in the discrete domain of the image coordinate space. This property makes Tchebichef moments superior to the conventional orthogonal moments such as Legendre moments and Zernike moments, in terms of preserving the analytical properties needed to ensure information redundancy in a moment set. The paper also details the various computational aspects of Tchebichef moments and demonstrates their feature representation capability using the method of image reconstruction.