Research

I successfully defended my PhD. thesis in computer science October 2017 at Toulon University, supervised by Frédéric Bouchara. My thesis is about the study of keypoints local features regarding applications in the document analysis and recognition field.

It is mainly about computer vision: I've tackled the issue of the inadequacy of keypoints detection algorithms (and local features) applied on document images.

Regarding this matter, my main contribution is a keypoint filtering algorithm in order to separate those prone to confusion from truly original ones. I named it CORE for COnfusion REduction algorithm. It relies on theory probability tools and kernel density estimation, without training.

Thus, my main research subjects include, but are not limited to: computer vision, document analysis and recognition, machine learning and GPU computation.

My current publications:

Journals

• Emilien Royer, Thibault Lelore, Frédéric Bouchara, COnfusion REduction (CORE) algorithm for local descriptors, floating-point and binary cases, Computer Vision and Image Understanding, Volume 158, May 2017, Pages 115-125, ISSN 1077-3142. http://www.sciencedirect.com/science/article/pii/S107731421730005X

Conferences

• Emilien Royer, Joseph Chazalon, Marçal Rusiñol and Frédéric Bouchara : Benchmarking Keypoint Filtering Approaches for Document Image Matching, international Conference on Document Analysis and Recognition (ICDAR) 2017 Best poster award
• Emilien Royer and Frédéric Bouchara : Guiding text image keypoints extraction through layout analysis, Camera Based Document Acquisition and Recognition (CBDAR) 2017.
• Emilien Royer - Thibault Lelore - Frédéric Bouchara : CORE : a COnfusion REduction algorithm for keypoints filtering, International Conference on Computer Vision Theory and Applications (VISAPP), march 2015