Image and Video Quality Assessment
Eye-Tracking in Visual Quality Assessment
Visual features drive our assessment of visual quality. Eye-Tracking studies into what parts of an image are relevant for a users decision on quality are important, to better understand how to improve quality and quality assessment. This project is a general collection for approaches to identify relevant visual features using eye-tracking.
Localization of Change-Salient Visual Features
Just noticeable difference describes the point at which human observers see changes made to visual media, e.g. through compression. However, the difference is initially only noticeable at a few salient locations. Finding the locations where users observe this change can lead to better encoding algorithms, by accounting for features of these change-salient regions. Therefore, this project aims to first locate these parts of the image and then derive features from them, to guide the image processing pipeline.
Automated Enhancement of Visual Features
Many factors play a role when users assess visual image quality or aesthetics. A lot of research in VQA has focused on identifying and assessing degradations and artifacts in a digital image, given a reference pristine image. However, reference images can also be enhanced to improve their visual features. This project aims to explore these enhancement methods of various visual aspects. Automated enhancement methods will be implemented and evaluated.
Improving Technical Quality Metrics
Conventionally, visual quality metrics rely on multiple subjective opinion scores that are aggregated into a mean opinion score or distributions of individual scores. Every human rater has their own personal biases, however, which influence their opinion. It would be desirable to debias the subjective scores, since this should yield a generalized and more exact mean opinion score, which in turn may increase performance of visual quality metrics. The aim of this project is to implement some forms of debiasing of subjective scores on existing VQA datasets and evaluating their impact on performance of prediction.
Devising Methods to Assess Small Visual Differences
Image quality has increased tremendously in the past years with new technological advancements. However, there are still minute differences in technical image quality caused by different parts of the image processing pipeline. Nonetheless, while the differences are small, it is still desirable to accurately distinguish their impact on image quality. For an untrained eye this may be a challenging task. Instead, one may amplify the artifacts caused by some part of the pipeling to enable e.g. crowd workers to perform this job accurately. The aim of this project is to devise approaches to perform artifact amplification, study the impact of the amplification, and validate, whether the amplification is valid in the sense that the quality relationship between the items is preserved.
Influence of Cadence on Optimal Pacing Strategies
In the Powerbike project we develop methods for data acquisition, analysis, modelling, optimisation and visualisation of performance parameters in endurance sports with emphasis on competitive cycling. Currently, one of our main research topics is the optimization of pacing strategies based on mathematical models for the cycling dynamics and the riders physiological parameters. In the starting phase and in very slow/fast segments the cadence will have an influence on these strategies due to the limited gear ratio. Therefore the restrictions due to cadence should be included in our optimization framework.Show more
Pedal Rotation Angle and Velocity Measurement Using an Arduino Microcontroller
In the Powerbike project we develop methods for data acquisition, analysis, modelling, optimisation and visualisation of performance parameters in endurance sports with emphasis on competitive cycling. Essential part of the project is a bicycle simulator based on a Cyclus2 ergometer and our own PC-based control software. The simulator enables to collect high-precision data and simulate real-world tracks in our lab. To improve the simulation we plan the integration of an arduino microcontroller with additional sensors e.g. optical or magnetic ones in order to better measure the rotation angle and velocity of pedals.Show more
The Role of Cadence on Oxygen Uptake Kinetics in Cycling
Respiratory gas exchange measurements are a common method to assess physical fitness and endurance capabilities of athletes. We develop models for predicting oxygen uptake kinetics with respect to a persons power output on a cycle ergometer. These models have successfully been validated with ergometer cycling exercises at a constant cadence.
However cycling exercise in the field is not restricted to constant cadences. Therefore some open questions remain. For example:
- Has a varying cadence an influence on the oxygen uptake and if so, to what extent?
- Does the maximal oxygen uptake change for different cadences?
- Finally, if there is an influence, how can we adjust our models appropriately?
Measuring and Modeling Wind during Road Cycling
During road cycling, one of the bigger uncertainties is wind. Wind direction and force both affect cycling performance.
One goal of the powerbike project, the optimization of pacing strategies for road cycling relies on a mathematical model for road cycling (Martin et al., 1998). The current state-of-the-art omits the possibility of having different wind forces and directions during a ride.
Therefore, we are looking forward for wind measurement solutions and an extended road cycling model.Show more
Between-day Variablity of VO2, VCO2 and Heart Rate for Variable Work Rate
Measurements of VO2, VCO2 and heart rate are central to methods for assessment of physical fitness and endurance capabilities in athletes. The aim of this study is to quantify the between-day variablity of VO2, VCO2 and heart rate measurements during ergometer tests with variable work rate.Show more