Development of an unsupervised inpainting methodology for occlusion removal in urban images

• Grant number: 18/06756-3
• Support Opportunities: Scholarships in Brazil - Scientific Initiation
• Start date: July 01, 2018
• End date: July 31, 2021
• Field of knowledge: Physical Sciences and Mathematics - Computer Science - Computing Methodologies and Techniques
• Principal Investigator: Wallace Correa de Oliveira Casaca
• Grantee: Dayara Pereira Basso
• Host Institution: Universidade Estadual Paulista (UNESP). Campus de Rosana. Rosana , SP, Brazil
• Associated scholarship(s): 19/24259-0 - Segmentation and inpainting of sensor failures in remote sensing images, BE.EP.IC

Abstract

Computer Vision (CV) techniques have contributed significantly to the development of several applications in the field of Remote Sensing. Although the progressive use of these computational apparatus has been shown useful for various broad tasks such as classification and feature extraction in aerial images, it is important noticing that, under certain circumstances, the efficiency of these algorithms tend to drastically decrease, leading to the impracticability of the desired goal. For example, in urban images, finding unwanted objects such as projections of building shadows, clouds, scenes containing street names, properties, etc whose overlapping content does not integrate the original imaging scenery has became very common in recent years. Such an issue makes the genuine data in the degraded regions inoperative for processing purposes by existing CV algorithms, impacting negatively in the accuracy of the results. In order to address the aforementioned occlusion problem, in particular for images containing textual content, building shadows, and other similar degradations, in this work we focus on investigating computer techniques for text/obstruction detection and object removal in aerial images. The proposed approach combines morphological operators for the task of occlusion detection and Partial Differential Equations to recover the detected area, thus generating a fully automatic methodology to properly address the object removal problem.

Scientific publications

(References retrieved automatically from Web of Science and SciELO through information on FAPESP grants and their corresponding numbers as mentioned in the publications by the authors)

BASSO, DAYARA; COLNAGO, MARILAINE; AZEVEDO, SAMARA; SILVA, ERIVALDO; PINA, PEDRO; CASACA, WALLACE. Combining morphological filtering, anisotropic diffusion and block-based data replication for automatically detecting and recovering unscanned gaps in remote sensing images. EARTH SCIENCE INFORMATICS, APR 2021. (18/06756-3, 19/24259-0)

BASSO, DAYARA; COLNAGO, MARILAINE; AZEVEDO, SAMARA; SILVA, ERIVALDO; PINA, PEDRO; CASACA, WALLACE. Combining morphological filtering, anisotropic diffusion and block-based data replication for automatically detecting and recovering unscanned gaps in remote sensing images. EARTH SCIENCE INFORMATICS, v. 14, n. 3, p. 14-pg., 2021-04-11. (18/06756-3, 19/24259-0)