Advanced search
Start date
Betweenand

Texture descriptors robust to rotation, lighting and colors

Grant number: 15/20812-5
Support type:Regular Research Grants
Duration: March 01, 2016 - February 28, 2018
Field of knowledge:Engineering - Electrical Engineering
Principal Investigator:Adilson Gonzaga
Grantee:Adilson Gonzaga
Home Institution: Escola de Engenharia de São Carlos (EESC). Universidade de São Paulo (USP). São Carlos , SP, Brazil
Assoc. researchers: Andrew Douglas Arnold Maidment ; Marcelo Andrade da Costa Vieira ; Predrag Bakic

Abstract

Texture analysis exerts a key role in computer vision. However, the lack of a mathematical definition and a consensus taxonomy among various researchers made texture definition be based most often on heuristics. In this line of thought, the most efficient and modern descriptor is the Local Binary Pattern (LBP). This approach has proven to be quite effective considering the sensitivity, precision and accuracy in segmentation and classification of texture in digital images. However, in uncontrolled environments, i.e. when the images or videos are acquired without lighting control or orientation of the intrinsic texture structures, most models fail and consume exorbitant computational time. Our proposal, based on previous search supported by FAPESP, is named Location Mapped Pattern (LMP), which is a actually a generalization of LBP and has generated more robust results than the original LBP. It is a parametric model that needs to be investigated further to produce more efficient results an to increase its performance. In this work, we propose the continuous development of LMP based on two actions: performance and computational time. In terms of performance, we propose developing descriptors to classify textures, with more accuracy than the results obtained by the LBP, both in mammographic images and in uncontrolled environments, where the texture could be rotated, colored and acquired with non-uniform illumination. In order to reduce the computational time, we propose to evaluate the reduction in size of the descriptors without loss of performance. (AU)

Scientific publications (4)
(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)
DE OLIVEIRA, HELDER C. R.; MENCATTINI, ARIANNA; CASTI, PAOLA; CATANI, JULIANA H.; DE BARROS, NESTOR; GONZAGA, ADILSON; MARTINELLI, EUGENIO; DA COSTA VIEIRA, MARCELO A. A cross-cutting approach for tracking architectural distortion locii on digital breast tomosynthesis slices. Biomedical Signal Processing and Control, v. 50, p. 92-102, APR 2019. Web of Science Citations: 0.
VIEIRA, RAISSA TAVARES; NEGRI, TAMIRIS TREVISAN; GONZAGA, ADILSON. Improving the classification of rotated images by adding the signal and magnitude information to a local texture descriptor. MULTIMEDIA TOOLS AND APPLICATIONS, v. 77, n. 23, p. 31041-31066, DEC 2018. Web of Science Citations: 1.
NEGRI, TAMIRIS TREVISAN; ZHOU, FANG; OBRADOVIC, ZORAN; GONZAGA, ADILSON. Extended color local mapped pattern for color texture classification under varying illumination. JOURNAL OF ELECTRONIC IMAGING, v. 27, n. 1 JAN 2018. Web of Science Citations: 0.
FERRAZ, CAROLINA TOLEDO; GONZAGA, ADILSON. Object classification using a local texture descriptor and a support vector machine. MULTIMEDIA TOOLS AND APPLICATIONS, v. 76, n. 20, p. 20609-20641, OCT 2017. Web of Science Citations: 0.

Please report errors in scientific publications list by writing to: cdi@fapesp.br.