Getting poster data...
A. Johannes Pretorius, Roy A. Ruddle (University of Leeds, Woodhouse Lane, Leeds, LS2 9JT, UK)Image analysis algorithms are often highly parameterized and they require significant human input to optimize input parameter settings. We propose a new approach based on parameter sampling and interactive visual exploration. Users are only involved in the first step - to initialize sampling - and the last step - to visually analyze the output in real time. To enable sampling, we have developed a custom plug-in for CellProfiler - a popular image analysis framework. After users have specified which parameters to sample and how, the plug-in automatically samples input parameters, computes and saves results to disk. Our main focus is on visual interfaces for analyzing and exploring relationships between sampled input parameters and corresponding output images. We have developed Paramorama - an interactive visualization tool - to facilitate this. It provides a visual overview of parameters and their sampled values and presents user-defined areas of interest in a structured fashion that incorporates scaled image results. It is highly interactive to enable users to identify and relate areas of interest in parameter and output space.