Complex Wavelet Features for Bark Texture Classification

Samuel Tschiedel GuedesDíbio Leandro Borges

Bark texture classification is a difficult image analysis problem because of the lack of regularity in those texture images. In this work bark texture features are extracted using magnitude coefficients of a Dual ­Tree Complex Wavelet Transform (DT­CWT). A database with acquired images of 51 classes of barks, 36 samples for each class, is tested computing recall and precision curves for 4 different distance metrics. Experiments are done also including Brodatz images and Gabor features. Results show that the DT­CWT coefficients are as accurate as Gabor features for bark texture classification if estimated and compared with equivalent orientations and scales. Advantages are to the DT­CWT features since they are faster to compute.

Caso o link acima esteja inválido, faça uma busca pelo texto completo na Web: Buscar na Web

Biblioteca Digital Brasileira de Computação - Contato:
     Mantida por: