Title: From shadows to illumination

Time and place: 1 pm in 3-228 (Las Vegas room)

The talk presents recent work in the area of detecting dynamic shadows in image sequences using a combination of color and depth information from the Bumblebee XB3 commercial stereo camera. The goal is to develop a technique that can run over very long sequences in outdoor scenarios with constant illumination changes. Therefore the technique is based on simple image differencing rather than relying on trained background images as is the current state-of-the-art. Dynamic shadow detection is demonstrated on a number of outdoor image sequences.
The talk also demonstrates how the detected dynamic shadows can be used for 1) bootstrapping a technique for detecting ALL shadows in the scene, and 2) estimating the illumination from the sky and from the sun.

PLANNED EVENTS

PAST EVENTS

Title: The Hermes Project

Time and place: 1 pm in 3-228 (Las Vegas room)

HERMES is an EU project wherein CVMT has participated in the last three years. The project concentrates on how to extract descriptions of peoples' behaviours from videos in a restricted discourse domain, such as pedestrians crossing inner-city roads, transform this into written text, and allow to synthesis a dynamic scene based on a textual description. The recorded videos allow to explore a coherent evaluation of human movements and facial expressions across a wide variation of scale. This general approach lends itself to various cognitive surveillance scenarios at varying degrees of resolution: from wide-field-of-view multiple-agent scenes, through to more specific inferences of emotional state that could be elicited from high resolution imagery of faces.

This talk presents the overall ideas behind HERMES together with concrete examples on what has been done in the project.

Title: Peach blossoms

PPT: Presentation

In spring I collected 32 GB of image data in the cold and moist Californian peach orchards. We had built transportable 3 camera stereo rig competing against time given by when the peach blossoms would bloom. Because of the winter rain season drought and delayed rain the blooming season was delayed 2-3 weeks, and the weather made it hard to actually work in the fields with expensive electronics in the fields. In this colloquium I will tell about all the joys of field work, new challenges in 3d stereo vision with 10Mpixel images, and preliminary results.

       

Title: Context-based adaptive filtering of interest points in image retrieval

Time and place: 1 pm in 3-228 (Las Vegas room)


Interest points have been used as local features with success in many computer vision applications such as image/video retrieval and object recognition. However, a major issue when using this approach is a large number of interest points detected from each image and created a dense feature space. This influences the processing speed in any run-time application. Selecting the most important features to reduce the size of the feature space will solve this problem. Thereby this raises a question of what makes a feature more important than the others? In this paper, we present a new technique to choose a subset of features. Our approach differs from others in a fact that selected feature is based on the context of the given image. Our experimental results show a significant reduction rate of features while preserving the retrieval performance.

       

September 28, 2009: Marcus Pearce, University College, London

Title: From expectation to aesthetics in the perception of music

Time and place: 1 pm in 3-228 (Las Vegas room)

Aesthetic accounts of music often tend towards one of two extremes: formalism and referentialism. Yet empirical research indicates that people value music largely because of the emotions that it evokes. Leonard Meyer pointed out that emotional expression, usually associated with referentialism, is actually compatible with both accounts and used expectation to reconcile formalism with emotional expression in music. Recent research in psychology and neuroscience has built on this approach studying the cognitive, physiological and neural components of expectation in music, covering both the factors that generate expectations and the effects of expectations on the listener. Researchers have identified three influences on a listener's expectations at a particular point in a novel piece of music: extra‐opus musical experience, intra‐opus structure and universal rules. I will argue that structural expectations are largely acquired and that rule‐like responses emerge as part of a process of implicit statistical learning. Turning to the effects of expectation, Meyer argued that confirmations and violations of expectation in music generate experiences of tension and resolution that in turn produce valenced emotional states. Again, there is empirical evidence that violations of expectation do indeed generate physiological signatures of emotion and activate reward circuits in the brain. Various explanations of how this process operates have been proposed. I suggest that it can be linked with the process of statistical learning and prediction that generates expectations in the first place.

September 16, 2009: Carsten Høilund

Title: Free Space Computation From Stochastic Occupancy Grids Based On Iconic Kalman Filtered Disparity Maps

Time and place: 1 pm in 3-228 (Las Vegas room)


The objective is to determine the free space in a scene as viewed by a camera on a vehicle. Free space is defined as the area where navigation without collision is possible. The starting point for determining this free space is disparity maps, obtained with a stereo camera. The disparity maps are filtered by an iconic Kalman filter, operating on each pixel individually. Applying ego-motion, the previous disparity maps is predicted to correspond to the current disparity map. The two, ideally identical, disparity maps are merged by the Kalman filter yielding an optimal estimation of the true state, thereby reducing variance and increasing the density of the filtered disparity map.

The stochastic occupancy grids are calculated from these disparity maps, providing a top-down view of the scene where the uncertainty of disparity measurements are taken into account. A pixel from the disparity map can thus affect several cells with varying likelihood.

These occupancy grids are segmented to indicate a maximum depth free of obstacles, enabling the marking of free space in the accompanying intensity image. The test shows successful marking of free space in the evaluated scenarios in addition to significant improvement in disparity map quality.

July 9, 2009: Martin Kraus (one of the applicants for our positions)

Title: What to Do with More Graphics Performance: Bigger, Better, Faster - or Something Different?

Time and place: 1 pm in room 3-108 (or in Las Vegas)

Graphics performance has been increasing dramatically during the last decade. Computer-animated movies are commonplace today and many researchers are working on real-time rendering with the same level of quality. This vision could be characterized as "bigger, better, faster". At the same time, the additional graphics performance also leverages very different approaches to many problems in computer graphics. Moreover, these new solutions also enable us to address new challenges.

This talk will sketch some of these new solutions and challenges that I've been working on in recent years. The topics include interactive illustrations, image interpolation, volume visualization and realistic rendering. I will try to show how more graphics performance not only changed our solutions but also the challenges that we are addressing.

May 6, 2009: Preben Fihl

Title: Computer Vision-Based Analysis of Human Actions

Time and place: 1 pm in room 3-108 (or in Las Vegas)

The talk will present an overview of the work I have done over the past four years within the area of analysis of human actions.

Video-based analysis of human motion has a wide range of interesting applications – from very detailed motion capture for movie production to coarse estimates of the flow of people in e.g. airports. Automatic analysis of surveillance video is another application which has received much interest in recent years, both from the political system, the industry and from the research community, and this talk will present work on different aspects of automatic analysis of surveillance video.

The talk will first present a method for foreground segmentation in relatively unconstrained outdoor scenes. Next I will present two different approaches to human action recognition. The first approach recognizes a set of one-arm gestures and the second approach classifies human gait into three types, i.e. walking, jogging, and running. The last part of the talk will present some ongoing work on full-body pose estimation from multiple cameras.

April 22, 2009: Anne-Marie S. Hansen

Title: Data logging for social aspects

Time and place: 1 pm in room 3-108 (or in Las Vegas)

Based on research within tangible embedded interaction, systematic musicology and music therapy, Anne-Marie's Ph.D. project explores new forms of playing, learning and socializing in an arbitrary physical space. In recent years it has become feasible to use wireless sensor technology in everyday life situations. This Ph.D. project investigates how users understand and interact with wireless interconnected sensor interfaces that provide reactive and adaptive sound feedback. The interfaces can be music instruments that adapt to the situation created by the users. It documents how multi-modal one-to-many, many-to-many interaction mappings cause interaction patterns that are measured qualitatively and quantitatively.


Anne-Marie Skriver Hansen will present the current stage of her Ph.D. research, which involves a continuous iterative loop that evolves around the industrial design of a tangible toy unit, the design and implementation of sound games, field studies of children's play and the analysis of the interaction patterns that arise from each sound game.

April 15, 2009: David Meredith (CANCELLED)

Title: Machine learning of musical style (or analysis and transcription of performed music)

April 1, 2009: IKT på Toppen (ICT on the Top)

Title: visit from representatives from the IKT på Toppen group

March 25, 2009: Claus B. Madsen

Title: On the analogies between computer graphics and classical 1D signal processing

    

There are several interesting analogies between modern computer graphics and classical 1D signal processing. Computing/determining the reflected radiance from a surface under some illumination conditions corresponds quite precisely to computing/determining the time-domain response of a filter to an input signal. The talk will "indicate" some of these analogies without making the presentation inaccessible to people with no signal processing background (hopefully). There will also be examples of how the big rotating arc in the lab can be/has been used to explore these analogies.

March 11, 2009: MED4 and MED6

Title: MED Awards spring 2009

March 4, 2009: Cornelius Poepel, Fachhochschule Ansbach, Germany

Title: Audio Signal Driven Sound Synthesis - ASDSS

I will present parts of my own research into Audio Signal Driven Sound Synthesis - ASDSS as well as the University and study course I come from. Since we have just set up the ERASMUS exchange program between the study programs Medialogy / Aalborg and Muiltimedia and Communication / Ansbach I will also talk about some classes/facilities as well as student works.

Research:
Field: HCI/Sound Synthesis
Topic: Audio Signal Driven Sound Synthesis

ERASMUS exchange Uni Aalborg / HS Ansbach
Study course Multimedia and Communication
Classes, Facilities, examples of students' works

Fachhochschule Ansbach