A DATA-DRIVEN APPROACH TO FEATURE SPACE SELECTION FOR ROBUST MICRO-ENDOSCOPIC IMAGE RECONSTRUCTION

In the article we propose a new on-line feature space selection strategy for displacement field estimation in the context of multi-view reconstruction of biological images acquired by a multi-photon micro-endoscope. While the high variety of targets encountered in clinical endoscopy induce enough texture feature variability to prohibit the use of recent supervised learning or feature matching based visual tracking methods, we will show how on-line learning combined with a classical method such as Digital Image Correlation (DIC) can contribute to the improvement of convex optimization-based template matching techniques.