DCVISIGRAPP 2014 Abstracts


Full Papers
Paper Nr: 1
Title:

A Novel Framework for Computing Unique People Count from Monocular Videos

Authors:

Satarupa Mukherjee and Nilanjan Ray

Abstract: Counting the unique number of people in a video (i.e., counting a person only once while the person is within the field of view), is required in many significant video analytic applications, such as transit passenger and pedestrian volume count in railway stations, malls and road intersections. The principal roadblock in this application is the occlusion. In my PhD thesis, we engineer a novel and straightforward solution to the problem by combining machine learning techniques with simple pixel motion tracking. We estimate the influx and/or the outflux rate of unique people in a region of interest within a monocular video. The unique count is then obtained by summing the influx and/or the outflux rates. Our proposed framework avoids people detection and people tracking that are plagued by occlusions. Also, it is online in nature without error accumulation so that unique people count can be obtained between any two time points in a streaming video. We validate the framework on 19 publicly available monocular videos. Occlusions are abundant in these videos, yet we obtain more than 95% accuracy for most of these videos. We also extend our proposed framework beyond monocular videos and apply it on multiple views of a publicly available dataset with about 99% accuracy.

Paper Nr: 2
Title:

Development of a Multispectral Gastroendoscope to Improve the Detection of Precancerous Lesions in Digestive Gastroendoscopy

Authors:

Sergio Ernesto Martinez Herrera, Yannick Benezeth, Matthieu Boffety, François Goudail, Dominique Lamarque, Jean-François Emile and Franck Marzani

Abstract: The diagnosis of malignancies in the stomach is based on the histological analysis and the visual information during gastroendoscopy. The visualization is mainly performed under white light; unfortunately, it is often difficult to visualize malignancies in gastric tissue using this technology, due to its subtle differences in colour and morphology in comparison with healthy tissues. Thus, it is clear that practitioners need additional information from the tissue in a non-invasive, efficient and accurate way. Gastric pathologies are related to small changes in the properties of the tissue which can be identified using multispectral imaging. The PhD thesis has two main objectives. The first one is the development of a gastroendoscopic prototype capable to acquire multispectral images of the stomach during gastroendoscopy. The second one is the proposal of tools and methods from the acquired spectra to identify cancerous tissue at an early stage; this is the current work for the second and third year of the PhD thesis.

Paper Nr: 3
Title:

Expressive Talking Head for Interactive Conversational Systems

Authors:

Paula Dornhofer Paro Costa and José Mario De Martino

Abstract: The synthesis of expressive speech videorealistic facial animation remains a challenging problem in computer graphics. The objective of this work is to propose a new synthesis methodology for an expressive talking head based on the manipulation of photographs, also referred as 2D facial animation. Our focus is directed to applications where the talking head act as an embodied conversational agent, with the ultimate goal of creating animated faces capable of inspiring user thrust and empathy.

Paper Nr: 5
Title:

Optimal Object Categorization under Application Specific Conditions

Authors:

Steven Puttemans and Toon Goedemé

Abstract: Day-to-day industrial computer vision applications focusing on object detection have the need of robust, fast and accurate object detection techniques. However, current state-of-the-art object categorization techniques only reach about 85% detection rate when performing in the wild detections who try to cope with as much scene and object variation as possible. However several industrial applications show many known characteristics like constant lighting, known camera position, constant background, … giving lead to several constraints on the actual algorithms. With a complete new universal object categorization framework, we want to prove the detection rate of these object categorization algorithms by exploiting the application specific knowledge which can help to reach a robust detector with detection rates of 99.9% or higher. We will use the same constraints to effectively reduce the number of false positive detections. Furthermore we will introduce an innovative active learning system based on this application specific knowledge that will drastically reduce the amount of positive and negative training samples, leading to a shorter and more effective annotation and training phase.

Paper Nr: 6
Title:

Estimating Driver Unawareness of Pedestrian based on Visual Behaviors and Driving Behaviors

Authors:

Minh Tien Phan, Indira Thouvenin, Vincent Fremont and Véronique Cherfaoui

Abstract: Taking into account the driver's state is a major challenge for designing new advanced driver assistance systems. In the context of pedestrian safety, a lot of research and many applications are able to detect the pedestrian with in-vehicle sensors and inform the driver of their presence. However, most of these alert systems do not adapt to the driver, they become distracting and are often ignored or deactivated. In this paper, we present a discussion on the danger to pedestrian that is related to the driver. We talk about the problem of pedestrian detectability. We do a review on the methods that help to detect the inattention state of the driver. Finally, we propose a new approach to analyze directly the driver’s awareness or unawareness of pedestrian based on driver’s visual and driving behaviors.

Paper Nr: 7
Title:

Multi-temporal Flow Maps - Looking Back to Look Forward

Authors:

Alberto Debiasi and Raffaele De Amicis

Abstract: Flow maps represent in a simple and clear way origin-destination data, mostly due to its simple design and intuitive understanding. In flow maps, aggregation techniques can be of valuable use because they reduce the visual clutter (e.g. each flow is easier to trace). The main objectiveof the research is to automatically generate flow maps and represent also multivariate attributes with temporal dimensions through interaction techniques, animations and the use of the 3D. Usually when there is the need to represent spatio-temporal data, the time dimension is simply associated to the z axis and this is a limit. A smart use of this dimension can open big opportunities to novel visualization techniques.

Paper Nr: 8
Title:

Hardware Implementation of Smart Embedded Vision Systems

Authors:

Elisa Calvo-Gallego, Piedad Brox and Santiago Sánchez-Solano

Abstract: The research presented in this contribution is focused on the efficient hardware implementation of image processing algorithms that are present at different levels of a smart vision system. The system is conceived as are configurable embedded device which, in turn, will be a node of a collaborative sensor network. The inclusion of fuzzy logic-based systems is explored to improve the performance of conventional vision algorithms.

Paper Nr: 9
Title:

Doctoral Thesis - A Visual Analysis System for Hierarchical Ensemble Data

Authors:

Matthias Thurau

Abstract: Analyzing ensemble data is very challenging due to the complexity of the task. In this paper, I describe IPFViewer, a visual analysis system for ensemble data, that is hierarchical, multidimensional, multivariate and multimodal. That system forms the basis for my doctoral thesis. An exemplary data set comes from a steel production facility and comprises data about their melting charges, samples and defects. My system differs from existing ones in that it encourages the usage of side-by-side visualization of ensemble members. Besides trend analysis, outlier detection and visual exploration, side-by-side visualization of detailed ensemble members enables rapid checking for repeatability of single ensemble member analysis results. IPFViewer supports the following data interaction methods: Hierarchical sorting and filtering, reference data selection, automatic percentile selection and ensemble member aggregation, while the focus for visualization is on small multiples of multiple views.

Paper Nr: 11
Title:

Computer-aided Diagnosis of Retinopathy of Prematurity via Analysis of the Vascular Architecture in Retinal Fundus Images of Preterm Infants

Authors:

Faraz Oloumi, Rangaraj M. Rangayyan and Anna L. Ells

Abstract: Retinopathy of prematurity (ROP) is a disorder that affects the development of blood vessels in the retina of premature infants, and is the leading cause of preventable childhood blindness. The posterior signs that are indicative of the presence of ROP are the straightening of the major temporal arcade (MTA), a decrease in the angle of insertion of the MTA, and increased dilation and tortuosity of the arteriole and venular vessels. Because advanced ROP can progress rapidly in the first 8 to 12 weeks of life, prompt identification of high-risk features of the disease is critical to the management of the affected infants.