Image Processing & Applications

Today, a wide range of domains encompassing, e.g., movie and video game production, virtual reality simulations, augmented reality applications, make a massive use of 3D computer generated assets. Although many graphics suites already offer a large set of tools and functionalities to manage the creation of such contents, they are usually characterized by a steep learning curve. This aspect could make it difficult for non-expert users to create 3D scenes for, e.g., sharing their ideas or for prototyping purposes. This paper presents a computer-based system that is able to generate a possible reconstruction of a 3D scene depicted in a 2D image, by inferring objects, materials, textures, lights, and camera required for rendering. The integration of the proposed system into a well-known graphics suite enables further refinements of the generated scene using traditional techniques. Moreover, the system allows the users to explore the scene into an immersive virtual environment for better understanding the current objects\' layout, and provides the possibility to convey emotions through specific aspects of the generated scene. The paper also reports the results of a user study that was carried out to evaluate the usability of the proposed system from different perspectives.

Keywords: Human-computer interaction , Image-based modeling , Scene and object modeling , Virtual reality

The current research reports for the first time the use of blends of poly(ε-caprolactone) (PCL) and poly(ester amide) (PEA) for the fabrication of 3D additive manufactured scaffolds. Tailor made PEA was synthesized to afford fully miscible blends of PCL and PEA using different percentages (5, 10, 15 and 20% w/w). Stability, characteristic temperatures and material's compatibility were studied through thermal analyses (i.e., TGA, DSC). Even though DMTA and static compression tests demonstrated the possibility to improve the storage modulus, Young's modulus and maximum stress by increasing the amount of PEA, a decrease of hardness was found beyond a threshold concentration of PEA as the lowest values were achieved for PCL/PEA (20% w/w) scaffolds (from 0.39 ± 0.03 GPa to 0.21 ± 0.02 GPa in the analysed load range). The scaffolds presented a controlled morphology and a fully interconnected network of internal channels. The water contact angle measurements showed a clear increase of hydrophilicity resulting from the addition of PEA. This result was further corroborated with the improved adhesion and proliferation of human mesenchymal stem cells (hMSCs). The presence of PEA also influenced the cell morphology. Better cell spreading and a much higher and homogenous number of cells were observed for PCL/PEA scaffolds when compared to PCL ones.

Keywords: Additive manufacturing , Biological properties , Image analysis , Poly(ester amide) , Scaffold design , Thermal and mechanical properties

The control of the process–structure–property relationship of a material plays an important role in the design of biomedical metal devices featuring desired properties. In the field of endodontics, several post-core systems have been considered, which include a wide range of industrially developed posts. Endodontists generally use posts characterized by different materials, sizes, and shapes. Computer-aided design (CAD) and finite element (FE) analysis were taken into account to provide further insight into the effect of the material–shape combination of metal posts on the mechanical behavior of endodontically treated anterior teeth. In particular, theoretical designs of metal posts with two different shapes (conical-tapered and conical-cylindrical) and consisting of materials with Young's moduli of 110 GPa and 200 GPa were proposed. A load of 100 N was applied on the palatal surface of the crown at 45◦ to the longitudinal axis of the tooth. Linear static analyses were performed with a non-failure condition. The results suggested the possibility to tailor the stress distribution along the metal posts and at the interface between the post and the surrounding structures, benefiting from an appropriate combination of a CAD-based approach and material selection. The obtained results could help to design metal posts that minimize stress concentrations.

Keywords: Computer-aided design (CAD) , Dental materials , Finite element analysis , Image analysis , Mechanical properties , Metal posts

This paper deals with parenthood perception (maternal and paternal) after the visualization and interaction (touch) with a 3D printed facial fetal model. The model is created using Additive Manufacturing techniques, starting from the image elaboration of routine ultrasound data. In this study, the method used for the elaboration and construction of 3D printable models of fetal faces starting from routine ultrasound images is briefly described. In addition, we present the results of a new survey conducted with future parents at the Altamedica clinic (Rome, Italy) to verify whether there are any benefits derived from the use of 3D printing models with future parents, both regarding the improvement of the parenthood experience, and the improvement of the understanding and collaboration with the physicians in case of fetal malformations, using 3D models coupled with the data of routine ultrasound examinations.

Keywords: 3D ultrasound , Additive manufacturing , Fetal face , Image processing , Parenthood perception , Survey

Low cost marker-less motion capture (Mocap) systems can be considered an interesting technology for the objective assessment of rehabilitation processes. In particular, this paper presents a feasibility study to introduce a Mocap system as a tool to assess shoulder rehabilitation. The movements of a shoulder are complex and challenging to be captured with a marker-less system because the skeleton avatar usually oversimplifies shoulder articulation with a single virtual joint. The designed solution integrates a low-cost Mocap system with image processing techniques and convolutional neural networks to automatically detect and measure potential compensatory movements executed during an abduction, which is one of the first post-surgery exercises for shoulder rehabilitation. First, we introduce the main steps of a reference roadmap that guided the development of the Mocap solution for rehab assessment of injured shoulder. Then, the acquisition of medical knowledge is presented as well as the new Mocap solution based on the integration of convolutional neural networks and 2D motion tracking techniques. Finally, the application which automatically evaluates abductions and makes available the measurements of the scapular elevations is described. Preliminary study and future works are also presented and discussed.

Keywords: Blender , Marker-less Motion capture system , Open source SDK , OpenPose , Shoulder rehabilitation

The rehabilitation process of patients after spinal cord injury (SCI) is usually based on subjective visual assessment by medical staff of rehabilitation centers. During the process, the medical personnel train patients to manage wheelchair and they learn how to use their sensible body parts in order to have a satisfactory life-style. Furthermore, physiotherapists and physicians have to control patients to prevent wrong postures that could cause further disorders. This paper describes how a low-cost marker-less motion capture system can be exploited to create an objective assessment procedure. Three Microsoft Kinect v2 sensors have been used to track patients using their wheelchairs along a straight path. The three sensors are arranged to optimize the acquisition. Thanks to the collaboration with the medical staff, we identified the set of parameters necessary to monitor patients' performance. An ad hoc application has been developed to provide the physicians with the right set of data easy readable to assess the patients along the rehabilitation process. The application has been tested involving twenty volunteers. Finally, results reached so far and further developments are summarized and discussed.

Keywords: Assessment of rehabilitation processes , Low-cost marker-less motion capture system , Microsoft Kinect v2 , Spinal cord injury

The use of fast and accurate scanning systems for human body digitization might pave the way towards the development of less invasive processes for medical manufacturing. In this work, an advanced measurement system for human body 3D reconstruction is used to design tailored assistive devices. The system is a photogrammetric 3D body scanner developed by the authors.

Keywords: 3D body measurements , assistive devices , medical manufacturing

Introduction and Objectives Fabrication processes for spinal orthoses require accurate three-dimensional (3D) models of the patients' trunk. Current methods for 3D reconstruction used in this field mainly include laser or structured light scanning; these methods are time expensive and invasive, especially for patients with partial disabilities. Therefore, a theoretically instant system for data acquisition of anatomical structure is highly desirable. The objective of this work is to show the feasibility of using digital photogrammetry for human body digitization to generate accurate 3D models of the patients' trunk for spinal orthoses fabrication. Materials and Methods Multiple synchronized two-dimensional images of the human torso are captured from different points of view using a photogrammetric scanner. A 3D model is generated using the state-of-the-art algorithms for point cloud and surface reconstruction. The digitized model is then used as input for the standard computer-aided design (CAD)/computer-aided manufacturing (CAM) process of fabrication. R4D from Rodin4D is used as prosthetics and orthotics CAD software. A robotic cell constituted by a six-axis KUKA KR 30-3 is used for milling a polyurethane foam. Vacuum forming is then adopted to generate the orthosis. Two spinal orthoses are fabricated using this approach and a classical one; then, they are evaluated using quantitative and qualitative metrics. Results The data acquisition using this approach lasts 50 milliseconds. The 3D reconstruction accuracy averages 0.21 ± 1.27 mm, which suits for the considered health care scenario. Results of the initial fitting of the orthoses fabricated with the presented method show better performances in terms of time (44%), product quality (35%), and patient experience (30%). Conclusions Digital photogrammetry can be used to enhance the data acquisition and data processing of anatomical surfaces for the CAD/CAM process of spinal orthoses. The data acquisition time, almost instant, allows an easy compliance of many patients. The data processing allows generating accurate models of the patient's body. The overall process generates orthoses with a better quality with respect to those manufactured using conventional procedures. ©

Keywords: CAD/CAM , fabrication techniques , photogrammetry , prosthetics and orthotics , spinal orthoses , three-dimensional reconstruction

A single-camera stereo-digital image correlation (stereo-DIC) system to obtain 3D full-field vibration measurements is proposed. The optical setup is composed of two planar mirrors and a single low frame rate camera, thus resulting in a compact and low-cost equipment. The two mirrors are used to create pseudo-stereo images of a target surface on the camera sensor, which are then correlated by using stereo-DIC. The image acquisition process is carried out at low frame rates and the Nyquist-Shannon frequency limitation is overcome by adopting a down-sampling approach under the hypothesis that the vibration signal is characterized by a single known frequency component. The developed pseudo-stereo DIC system allows to obtain 3D full-field vibration measurements in a frequency range up to 4 kHz even with an available frame rate (at full resolution) of 178 fps. The effectiveness of the described approach has been verified by performing vibration measurements on a cantilever plate and a turbine blade.

Keywords: Digital image correlation , Down-sampling approach , Single low-speed camera , Vibration measurement

The concept of "surface modeling" generally describes the process of representing a physical or artificial surface by a geometric model, namely a mathematical expression. Among the existing techniques applied for the characterization of a surface, terrain modeling relates to the representation of the natural surface of the Earth. Cartographic terrain or relief models as threedimensional representations of a part of the Earth's surface convey an immediate and direct impression of a landscape and are much easier to understand than two-dimensional models. This paper addresses a major problem in complex surface modeling and evaluation consisting in the characterization of their topography and comparison among different textures, which can be relevant in different areas of research. A new algorithm is presented that allows calculating the fractal dimension of images of complex surfaces. The method is used to characterize different surfaces and compare their characteristics. The proposed new mathematical method computes the fractal dimension of the 3D space with the average space component of Hurst exponent H, while the estimated fractal dimension is used to evaluate, compare and characterize complex surfaces that are relevant in different areas of research. Various surfaces with both methods were analyzed and the results were compared. The study confirms that with known coordinates of a surface, it is possible to describe its complex structure. The estimated fractal dimension is proved to be an ideal tool for measuring the complexity of the various surfaces considered.

Keywords: Fractal dimension , Hurst exponent H , Image analysis , Space component , Surface

Purpose: The study aims to investigate the correlations between positron emission tomography (PET) texture features, X-ray computed tomography (CT) texture features, and histological subtypes in non-small-cell lung cancer evaluated with 2-deoxy-2-[18F]fluoro-D-glucose PET/CT. Procedures: We retrospectively evaluated the baseline PET/CT scans of 81 patients with histologically proven non-small-cell lung cancer. Feature extraction and statistical analysis were carried out on the Matlab platform (MathWorks, Natick, USA). Results: Intra-CT correlation analysis revealed a strong positive correlation between volume of the lesion (CTvol) and maximum density (CTmax), and between kurtosis (CTkrt) and maximum density (CTmax). A moderate positive correlation was found between volume (CTvol) and average density (CTmean), and between kurtosis (CTkrt) and average density (CTmean). Intra-PET analysis identified a strong positive correlation between the radiotracer uptake (SUVmax, SUVmean) and its degree of variability/disorder throughout the lesion (SUVstd, SUVent). Conversely, there was a strong negative correlation between the uptake (SUVmax, SUVmean) and its degree of uniformity (SUVuni). There was a positive moderate correlation between the metabolic tumor volume (MTV) and radiotracer uptake (SUVmax, SUVmean). Inter (PET-CT) correlation analysis identified a very strong positive correlation between the volume of the lesion at CT (CTvol) and the metabolic volume (MTV), a moderate positive correlation between average tissue density (CTmean) and radiotracer uptake (SUVmax, SUVmean), and between kurtosis at CT (CTkrt) and metabolic tumor volume (MTV). Squamous cell carcinomas had larger volume higher uptake, stronger PET variability and lower uniformity than the other subtypes. By contrast, adenocarcinomas exhibited significantly lower uptake, lower variability and higher uniformity than the other subtypes. Conclusions: Significant associations emerged between PET features, CT features, and histological type in NSCLC. Texture analysis on PET/CT shows potential to differentiate between histological types in patients with non-small-cell lung cancer.

Keywords: Non-small-cell lung cancer , Radiomics , Texture analysis , [ F] FDG PET/CT 18

Convolutional Neural Networks (CNN) have brought spectacular improvements in several fields of machine vision including object, scene and face recognition. Nonetheless, the impact of this new paradigm on the classification of fine-grained images-such as colour textures-is still controversial. In this work, we evaluate the effectiveness of traditional, hand-crafted descriptors against off-the-shelf CNN-based features for the classification of different types of colour textures under a range of imaging conditions. The study covers 68 image descriptors (35 hand-crafted and 33 CNN-based) and 46 compilations of 23 colour texture datasets divided into 10 experimental conditions. On average, the results indicate a marked superiority of deep networks, particularly with non-stationary textures and in the presence of multiple changes in the acquisition conditions. By contrast, hand-crafted descriptors were better at discriminating stationary textures under steady imaging conditions and proved more robust than CNN-based features to image rotation.

Keywords: Colour texture , Convolutional neural networks , Feature extraction , Hand-crafted image descriptors , Image classification

Filtering has been one of the main approaches to texture analysis since early on. Traditionally, the process involved designing the filters essentially by hand based on some prior knowledge (e.g. perceptual models, optimal mathematical properties, etc.) In this work we propose the use of convolutional networks for refactoring traditional, hand-designed filters. Our method consists of initialising the first convolutional layer of the network with some classic banks of filters, training the network on texture images and retrieve the modified filters. Experimenting with five classes of filters and eight datasets of texture images we show that the refactored filters can be conveniently used ‘off-the-shelf' to achieve better performance than obtained with the original filters, but at the same computational cost.

Keywords: Convolutional neural networks , Image filters , Texture analysis

This paper compares the effects of colour pre-processing on the classification performance of H&E-stained images. Variations in the tissue preparation procedures, acquisition systems, stain conditions and reagents are all source of artifacts that can affect negatively computer-based classification. Pre-processing methods such as colour constancy, transfer and deconvolution have been proposed to compensate the artifacts. In this paper we compare quantitatively the combined effect of six colour pre-processing procedures and 12 colour texture descriptors on patch-based classification of H&E-stained images. We found that colour pre-processing had negative effects on accuracy in most cases – particularly when used with colour descriptors. However, some pre-processing procedures proved beneficial when employed in conjunction with classic texture descriptors such as co-occurrence matrices, Gabor filters and Local Binary Patterns.

Keywords: Colour , Eosin , Hematoxylin , Histology , Texture

In this paper we investigate extensions of Local Binary Patterns (LBP), Improved Local Binary Patterns (ILBP) and Extended Local Binary Patterns (ELBP) to colour textures via two different strategies: intra-/inter-channel features and colour orderings. We experimentally evaluate the proposed methods over 15 datasets of general and biomedical colour textures. Intra- and inter-channel features from the RGB space emerged as the best descriptors and we found that the best accuracy was achieved by combining multi-resolution intra-channel features with single-resolution inter-channel features.

Keywords: Colour , Local binary patterns , Texture

Despite the fact that image descriptors based on the statistical distribution of local patterns are very common tools, most of their mathematical underpinnings have been largely overlooked. Among them, the rigorous determination of the number of possible patterns that can arise from a given neighbourhood and kernel function – particularly when invariance under group actions (e.g., rotations and/or reflections) is taken into account – has received little or no attention in the literature. In this note we address the problem of counting local patterns in a rigorous way. We provide exact formulas for the number of the possible directional, rotation- and reflection-invariant patterns generated by neighbours of n points over alphabets of k symbols. Variations on this scheme such as rank and uniform patterns are considered, and direct applications to a number of common descriptors (e.g., Local Binary Patterns, Texture Spectrum and Full Ranking) are also presented.

Keywords: Bracelets , Local image descriptors , Necklaces , Patterns

Wear and rolling contact fatigue are competing phenomena in railway wheels, as wear tends to shorten or remove surface cracks nucleated by ratcheting. The presence of water at the contact interface can enhance crack propagation leading to fatigue failure. This topic was studied taking advantage of innovative measurement systems developed for assessing the damage in bi-disc rolling contact tests, including a vision system for the acquisition and elaboration of surface images and a machine-learning technique for vibration measurement and analysis. Tests of different total duration with alternated dry and wet contact phases were carried out. The analysis of the collected measurements allowed identifying when crack propagation begins to prevail on wear: this occurred well earlier than the visible emergence of fatigue damage. If short dry and wet contact sessions are alternated, the onset of fluid driven crack propagation is delayed, because initially the dry sessions are not long enough to allow surface cracks to form by ratcheting, and in the subsequent wet session ratcheting is suspended due to low friction. If the alternated dry-wet contact sessions are longer, the onset of fluid driven crack propagation is accelerated, as in the dry sessions ratcheting proceeds more forming longer surface cracks, which are able to propagate in the subsequent wet phase.

Keywords: Crack propagation , Dry-wet contact , Ratcheting , Vibration analysis , Vision systems , Wheel steel

The paper describes the design of a wearable and wireless system that allows the real-time identification of some gestures performed by basketball players. This system is specifically designed as a support for coaches to track the activity of two or more players simultaneously. Each wearable device is composed of two separate units, positioned on the wrists of the user, connected to a personal computer (PC) via Bluetooth. Each unit comprises a triaxial accelerometer and gyroscope, a microcontroller, installed on a TinyDuino platform, and a battery. The concept of activity recognition chain is investigated and used as a reference for the gesture recognition process. A sliding window allows the system to extract relevant features from the incoming data streams: mean values, standard deviations, maximum values, minimum values, energy, and correlations between homologous axes are calculated to identify and differentiate the performed actions. Machine learning algorithms are implemented to handle the recognition phase.

Performance characteristics of the products made of metallic materials such as wear resistance, fatigue strength, stability of gaps and strain between the connections, corrosion resistance, etc., depend to a large extent by the quality of their surfaces roughness. An interactive control of the manufacturing parameters which influence the surface roughness is particularly crucial in the construction of many mechanical components. The present paper devises a new method for statistical pattern recognition on samples produced by the process of robot laser hardening using network theory and describes its application to the determination of surface roughness. The method is based on the analysis of SEM images. Indeed the data characterizing the state of surface irregularities detected as extremely small segments contain indicators of surface roughness. Different methods of machine learning techniques designed to predict the surface roughness of robot laser hardened material are discussed.

Keywords: Machine interactive learning , Robot laser hardening , SEM images , Statistical pattern recognition , Surface roughness