Localized matching using earth mover’s distance towards discovery of common patterns from small image samples

This paper proposes a new approach for the discovery of common patterns in a small set of images by region matching. The issues in feature robustness, match-ing robustness and noise artifact are addressed to delve into the potential of using regions as the basic matching unit. We novelly employ the many-to-many (M2M) matching strategy, specifically with the Earth Mover’s Distance (EMD), to increase resilience towards the structural inconsistency from improper region segmentation. However, the matching pattern of M2M is dispersed and unregulated in nature, lead-ing to the challenges of mining a common pattern while identifying the underlying transformation. To avoid analysis on unregulated matching, we propose localized matching for the collaborative mining of common patterns from multiple images. The patterns are refined iteratively using the expectation-maximization algorithm by taking advantage of the ‘crowding ’ phenomenon in the EMD flows. Experimen-tal results show that our approach can handle images with significant image noise and background clutter. To pinpoint the potential of Common Pattern Discovery (CPD), we further use image retrieval as an example to show the application of CPD for pattern learning in relevance feedback