News |  Sitemap |  Contact
PDF Export  | 

Relative-entropy based distance for automated detection of embryo in X-ray images of dry seeds of sugarbeet with ImageJ software

Abstract

In this report, we present an original image processing pipeline that we have developed under the ImageJ software for a specific application in the domain of plant science. In this domain, numerical X-ray imaging is applied more and more often for its capabilities for noninvasive inspection of internal structures of interest, in order to assess, for instance, the germinative quality of dry seeds [1]. Here, we address the task of detecting the presence or absence of an embryo in dry seed of sugarbeet. Each X-ray image can contain from ten to one hundred seeds. Our image processing pipeline first performs a segmentation of each seed, and constructs the histogram of the gray levels of each segmented seed. It then calculates a relative-entropy based distance, or Kullback-Leibler distance [2] between the normalized histogram of each seed and the histogram of a seed of reference with embryo and a seed of reference without embryo. The smallest Kullback-Leibler distance is chosen to decide the content of the seed under test. We will present in the extended version of the report, the confrontation of the ImageJ pipeline for automated detection, with the detection performed from visual inspection by a human expert. The pipeline, fully automated, can contribute to high-throughput screening and phenotyping of large quantities of seeds.

[1] E. Belin, D. Rousseau, J. Léchappé, M. Langlois-Meurinne, C. Dürr, “Rate-distorsion tradeoff to optimize high-throughput phenotyping systems. Application to X-ray images of seeds”, Computers and Electronics in Agriculture, vol. 77, pp. 188-194 (2011).

[2] T. M. Cover, J. A. Thomas, “Elements of Information Theory”, Wiley, 2006.

This work received funding from the ANR in the framework of the AKER “Investissements d'avenir” project.

Keywords

Image processing task, relative-entropy based distance, automated detection, X-ray images, dry seeds of sugar beet

Short CV

Etienne BELIN is associate professor with Université d'Angers. His research interests include image processing and its application to plant imaging. He uses ImageJ for interaction with life science scientists and for educational purposes.

Administrative data

Presenting author: Etienne BELIN
Organisation: Laboratoire d'Ingénierie des Systèmes Automatisés (LISA) - Université d'Angers

co-authors: David ROUSSEAU, Didier DEMILLY, Karima BOUDEHRI, François CHAPEAU-BLONDEAU, Carolyne DURR

© Luxembourg Institute of Science and Technology | Legal Notice