Visualization of Parameter Space for Image Analysis

Abstract

Image analysis algorithms are often highly parameterized and much human input is needed to optimize parameter settings. This incurs a time cost of up to several days. We analyze and characterize the conventional parameter optimization process for image analysis and formulate user requirements. With this as input, we propose a change in paradigm by optimizing parameters based on parameter sampling and interactive visual exploration. To save time and reduce memory load, users are only involved in the first step - initialization of sampling - and the last step - visual analysis of output. This helps users to more thoroughly explore the parameter space and produce higher quality results. We describe a custom sampling plug-in we developed for CellProfiler - a popular biomedical image analysis framework. Our main focus is the development of an interactive visualization technique that enables users to analyze the relationships between sampled input parameters and corresponding output. We implemented this in a prototype called Paramorama. It provides users with a visual overview of parameters and their sampled values. User-defined areas of interest are presented in a structured way that includes image-based output and a novel layout algorithm. To find optimal parameter settings, users can tag high- and low-quality results to refine their search. We include two case studies to illustrate the utility of this approach.

Facing the high-dimensions: Inverse projection with radial basis functions

“…Going from surprise to fear (points 1-4), fear to neutral (5-7), neutral to angry (8 and 9), angry to happy (10)(11)(12) happy to disgust (13 and 14) disgust to sad (15 and 16) and sad to surprise (17)(18)(19)(20). The faces generated in points 3 and 4 are good examples of blending of two expressions, namely surprise and fear.…”

“…Some systems, like the ones presented by Bruckner and Möller [16] and Pretorius et al [17] have also been designed to assist the user to explore multidimensional parameter spaces. In the paper of Bruckner et al [16], it was introduced a system that allows the user to explore simulation results for special effects.…”

“…Therefore, the system is not designed to support resampling of the parameter space. Likewise, in the paper of Pretorius et al [17] the authors proposed an interactive visualization technique that enables users to analyze the relationship between sampled input parameters and corresponding outputs. Their system is designed to assist users in the task of parameter tuning for image analysis.…”

Parameter Space Visualization for Large-scale Datasets Using Parallel Coordinate Plots

“…5 Siirtola has introduced two browser-based techniques for manipulating parallel coordinate plots. 6 The first technique uses polyline averaging to summarize a set of polylines. The second provides a visualization for correlation coefficients between polyline subsets in order to help the user to discover new information.…”

“…In this work, the paradigm of parameter sampling is changed in order to incorporate large parameter sweeps in a more efficient way. 6 The proposed system extends this methodology by making parameter sweeps using multiple models and comparing their outputs through a parallel coordinate visualization. Zhou et al developed several web-based visualization frameworks combined with preprocessing tools to provide a way for domain specialists to interpret their data.…”