Region Growing for Segmenting Green Microalgae Images

Author(s)

Borges, Vinicius R. P. ^[1] ; de Oliveira, Maria Cristina F. ^[1] ; Silva, Thais Garcia ^[2] ; Henriques Vieira, Armando Augusto ^[2] ; Hamann, Bernd ^[3]

Affiliation

^[1] Univ Sao Paulo, Inst Ciencias Matemat & Computacao, BR-13565905 Sao Carlos, SP - Brazil

^[2] Univ Fed Sao Carlos, Dept Bot, BR-13565905 Sao Carlos, SP - Brazil

^[3] Univ Calif Davis, Dept Comp Sci, Davis, CA 95616 - USA

• Document type: Journal article
• Source: IEEE-ACM TRANSACTIONS ON COMPUTATIONAL BIOLOGY AND BIOINFORMATICS; v. 15, n. 1, p. 257-270, JAN-FEB 2018.
• Web of Science Citations: 1

Abstract

We describe a specialized methodology for segmenting 2D microscopy digital images of freshwater green microalgae. The goal is to obtain representative algae shapes to extract morphological features to be employed in a posterior step of taxonomical classification of the species. The proposed methodology relies on the seeded region growing principle and on a fine-tuned filtering preprocessing stage to smooth the input image. A contrast enhancement process then takes place to highlight algae regions on a binary pre-segmentation image. This binary image is also employed to determine where to place the seed points and to estimate the statistical probability distributions that characterize the target regions, i.e., the algae areas and the background, respectively. These preliminary stages produce the required information to set the homogeneity criterion for region growing. We evaluate the proposed methodology by comparing its resulting segmentations with a set of corresponding ground-truth segmentations (provided by an expert biologist) and also with segmentations obtained with existing strategies. The experimental results show that our solution achieves highly accurate segmentation rates with greater efficiency, as compared with the performance of standard segmentation approaches and with an alternative previous solution, based on level-sets, also specialized to handle this particular problem. (AU)

• FAPESP's process: 11/22749-8 - Challenges in exploratory visualization of multidimensional data: paradigms, scalability and applications
• Grantee: Luis Gustavo Nonato
• Support Opportunities: Research Projects - Thematic Grants
• FAPESP's process: 13/26647-0 - Visual exploration to support green algae taxonomic classification
• Grantee: Vinícius Ruela Pereira Borges
• Support Opportunities: Scholarships abroad - Research Internship - Doctorate
• FAPESP's process: 12/00269-7 - Visual exploration of feature spaces to support Green Algae Taxonomic Classification
• Grantee: Vinícius Ruela Pereira Borges
• Support Opportunities: Scholarships in Brazil - Doctorate