Retinal Image Multimodal and Multitemporal Registration for a Multispot Laser Photocoagulation Device

• Published in: 2019 XV Workshop de Visão Computacional (WVC) pp. 21 - 25
• Main Authors: Patricia, Otoni P., Evandro, Rodrigues L. L.
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.09.2019
• Subjects: Diseases; Feature extraction; Histograms; Image color analysis; Image registration; Laser theory; local feature; multimodal registration; multispot laser; partial intensity invariance; PIIFD; Retina; retina images; robust point machine; SIFT; SURF; VFC
• DOI: 10.1109/WVC.2019.8876918

This paper aims to assist the physician during the treatment of retinal diseases by laser photocoagulation, aligning images obtained by different acquisition methods in preliminary examinations to the treatment. Thus, an algorithm with combinations and arrangements of related methods was proposed, in order to allow the multimodal recording of retinal images. The proposed algorithm performs the detection of points of interest using SURF, elaborates the characteristic descriptor of each point using a combination of SIFT and PIIFD, eliminates the outliers by means of a Robust Point Matching method via Vector Field Consensus (VFC) and finally performs the registration with the Geometric Transformation of the type Affine. The algorithm was tested in 40 pairs of multimodal retinal images and performed better when compared to existing algorithms in terms of precision and robustness.