Advanced search
Start date

Evaluation of color descriptors in natural images at various resolutions aided by a visualization tool

Grant number: 10/19159-1
Support type:Scholarships in Brazil - Scientific Initiation
Effective date (Start): March 01, 2011
Effective date (End): February 29, 2012
Field of knowledge:Physical Sciences and Mathematics - Computer Science - Computing Methodologies and Techniques
Principal Investigator:Moacir Antonelli Ponti
Grantee:Camila Tatiana Picon
Home Institution: Instituto de Ciências Matemáticas e de Computação (ICMC). Universidade de São Paulo (USP). São Carlos , SP, Brazil


Color descriptors are widely used in image pattern recognition, especially in applications of computer vision and content-based image retrieval. It is important to use descriptors with invariance to illumination, affine transformations, occlusion and that are robust to noise. Some studies were performed to determine and confirm these properties. This project aims the development and performance evaluation of color image descriptors when processed from images of various resolutions. Two databases of natural images are used to evaluate the performance of each type of color descriptor with images of different resolutions. The results are analysed by confusion matrices and accuracy rate, aided by a visualization tool to assist the image classification tasks. The results obtained can contribute on the selection of lower resolutions of images for various applications and a consequent improvement in the computational cost.