Anca Morar

Vessel visualization using curvicircular feature aggregation

Radiological investigations are common medical practice for the diagnosis of peripheral vascular diseases. Existing visualization methods such as Curved Planar Reformation (CPR) depict calcifications on vessel walls to determine if blood is still able to flow. While it is possible with conventional CPR methods to examine the whole vessel lumen by rotating around the centerline of a vessel, we propose Curvicircular Feature Aggregation (CFA), which aggregates these rotated images into a single view. By eliminating the need for rotation, vessels can be investigated by inspecting only one image. This method can be used as a guidance and visual analysis tool for treatment planning. We present applications of this technique in the medical domain and give feedback from radiologists.

From HTML to 3DMMO – A Roadmap Full of Challenges

WWW is everywhere, embraced by humankind to share information and collaborate. The basic schema of hyperlinked web pages has evolved with revolutionary technologies, altogether with the devices used to access the web, but still carries on strong limitations. Virtual reality and 3D MMO, with their sheer power of representation and interaction, is seen by many as the future of human computer based interaction and collaboration and would, naturally, mix with the WWW model sooner or later. The complexity of the technology, from technical point of view but also with many interdisciplinary issues (medical, ethical, economical, socio-political etc.) would make this transition challenging and exciting. This paper classifies, identifies and proposes basic conceptual solution for these issues - of moving from interconnected web pages to interconnected 3D virtual spaces with massive number of interacting users.

Kinect depth inpainting in real time

RGBD cameras are used in many fields, such as monitoring, robot navigation and assistive solutions for visually impaired people. The advance in technology led to the development of cheap and easy to use cameras, such as Kinect, that gained a lot of popularity. One of the technologies that allow the capturing of depth maps, the D component in RGBD sequences, is structured light. The main drawback of this technology is the noise introduced during the acquisition process and the missing values for many pixels in the depth map. This paper presents several new depth inpainting algorithms and their results. The algorithms improve depth signal significantly and can be used as a pre-processing step for segmentation or reconstruction of RGBD images based on depth maps.

Multi GPGPU Optimizations for 3D MMO Virtual Spaces

The constantly growing demand for 3D Massively Multiuser Online (MMO) virtual spaces has led to an increase of the load that current 3D MMO server architectures need to cope with in order to accommodate a large user base that interacts inside the virtual space at a given moment of time. The traditional methods that are employed to solve this computational challenge by the current 3D MMO server architectures have inherent drawbacks regarding scalability, reliability, redundancy, cost and the realism of the virtual world. In the last few years, the graphics hardware has evolved in a spectacular manner in terms of computing power and architectural design, switching from a “single-core” to current “many-core” architecture. In this paper we propose an innovative approach to tackle the heavy computational operations that are executed by the simulations of 3D MMO Virtual Spaces using Multi GPGPU (General Purpose programming on Graphical Processing Units).

Time-consistent segmentation of indoor depth video frames

We propose a new method for segmenting indoor scenes captured with RGBD or depth cameras with a simple and fast iterative CPU segmentation. The segmentation consists of three main steps. In the first stage, regions grow by adding new pixels based on depth information and normal vectors. Next, multiple regions are merged into bigger surfaces based on adjacency and region statistics. The last stage of the method ensures interframe consistency based on a cost function between regions from subsequent frames. The planar segmentation can be used both for identifying indoor objects, which in general have planar faces, as well as for estimating the camera motion and then reconstructing the environment.

Perceptual feedback training for improving spatial acuity and resolving front-back confusion errors in virtual auditory environments

This paper aims to test the hypothesis that audiovisual perceptual feedback based training can improve the sound localization accuracy and the navigational skills in a virtual auditory display where 3D sounds are created using non-individualized HRTFs. In our experiment, we tested the spatial auditory localization performance of nine sighted subjects prior and consequently a series of training procedures based on audiovisual feedback. The results concluded that this cross-modal facilitation enhances spatial hearing, reduces the incidence of localization and front-back confusion errors and helps the users to navigate more accurately in the virtual auditory environments by improving their ability to detect 3D sound sources synthesized with non-individualized HRTFs.

Real Time Indoor 3D Pipeline for an Advanced Sensory Substitution Device

In this paper, we present the indoor 3D pipeline of an assistive system for visually impaired people, whose goal is to scan the environment, extract information of interest and send it to the user through haptics and sounds. The particularities of indoor scenes, containing man-made objects, with many planar faces, led us to the idea of developing the 3D object recognition algorithms around a planar segmentation, based on normal vectors. The 3D pipeline starts with acquiring depth frames from a range camera and synchronized IMU data from an inertial sensor. The pre-processing stage computes normal vectors in the 3D points of the scanned environment and filters them to reduce the noise from the input data. The next stages are planar segmentation and object labeling, which divides the scene into ground, ceiling, walls and generic objects. The whole 3D pipeline works in real-time on a consumer laptop at approximately 15 fps. We describe each step of the pipeline, with the focus on the labeling stage, and present experimental results and ideas for further improvements.

Near Infrared Imaging for Tissue Analysis

Near infrared light or NIR is not absorbed well by water or blood, which are the main components of biological tissue. For this reason, it can reach a significant depth of optical penetration, thus achieving increased illumination of internal structures. Even if there are some approaches based on near infrared light that allow the analysis of tissue near the skin, there is very little research regarding the 3D reconstruction and visualization of NIR illuminated areas. The current paper presents a review of existing work, analyses the drawbacks and advantages of different methods and proposes new approaches to improve this imaging modality.

Comparison between Incompressible SPH Solvers

In this paper, we present two high performing and efficient Smoothed Particle Hydrodynamics (SPH) methods used to physically simulate incompressible fluids: Implicit Incompressible SPH (IISPH) and Divergence-Free SPH (DFSPH). At the moment, the most stable SPH methods employ density correctors that reduce the density error of the fluid particles to achieve very low levels of compressibility. We analyze the performance of the correctors used for the methods previously mentioned. We also compare IISPH against DFSPH to study several characteristics such as the stability of the particles, the compressibility levels of the fluid achieved during the simulations, the execution speed of each corrector, the maximum possible increase of the time steps, etc.

GPU accelerated 2D and 3D image processing

The current advances in hardware led to the development of the GPGPU (General-purpose computing on graphics processing units) paradigm. Thus, nowadays, the GPU (Graphics Processing Unit) is used not only for graphics programming, but also for general purpose algorithms. This paper discusses several methods regarding the use of CUDA (Compute Unified Device Architecture) for 2D and 3D image processing techniques. Some general rules for writing parallel algorithms in computer vision are pointed out. A theoretic comparison between the complexity for CPU (Central Processing Unit) and GPU implementations of image processing algorithms is given. Also, real computing times are provided for several algorithms in order to point out the actual performance gain of using the GPU over CPU. The factors that contribute to the difference between theoretic and real performance gain are also discussed.