Data Mining for Image and Video Analysis

Data Mining for Image and Video Analysis

The extraction of relevant visual patterns in images and videos are essential to characterize the structure of the object considered. These patterns can be used to define new representations, to propose some data summarizations/characterizations or to detect some particular elements in images or videos or even perform some classification or tracking tasks. We are particularly intested in the extraction of subsequences of graph-based patterns for object tracking in videos.


The extraction of relevant visual patterns in images or videos is a key element for building new representations or data summarization, event detection,  classification or visual tracking tasks. In this are we propose to extract patterns coming from (sub)graph representation when graph models are use to encode images or videos. One of our perspective is to work on the integration of probabilistic information in the patterns extracted.

In this area, our research work includes:

-Extraction of frequent subgraph patterns.We work on the definition of new extraction methods for particular types of graphs to help object tracking in videos. This approach is a natural follow-up of a previous works about the extraction of subgraph patterns from planar graphs. Our aim is to make use of subsequences of extracted patterns to improve tracking.

-Incorporation of probabilistic information in extracted patterns. We study the interest of using some probabilistic information - coming from probabilistic models used as an a priori knowledge for the considered task - into the extracted patterns. The objective is to mine more flexible patterns with better invariance and discriminative power.

-Extraction of temporal patterns. We are also interested in extraction temporal patterns in videos. This topic is also addressed in the Machine Learning for Computing Vision Applications theme.


previous topic: Data Mining for Structured Data                                    next topic: Social and Personalized Information Retrieval