K-NN based automated reasoning using bilateral filter based texture descriptor for computing texture classification

doi:10.1016/j.eij.2018.01.003

Return

K-NN based automated reasoning using bilateral filter based texture descriptor for computing texture classification

2018-07-01

Sonali Dash

Manas Ranjan Senapati *

Uma Ranjan Jena

DOI：10.1016/j.eij.2018.01.003

Original

Save

View PDF

Abstract

En 中文

Regions in the visual field can be characterized by differences in texture, brightness, colour, or other attributes. Bilateral filter is an efficient way to smooth any digital image while preserving the fine information. In bilateral filter, it has been observed that by selecting carefully, the bilateral filter range parameter and bilateral filter domain parameter the ability to smooth any arbitrary digital image while preserving the edges can be improved. This trait of bilateral filter helps to adapt it to application specific requirements. In this study, a new feature extraction method is recommended by integrating the conventional Laws' mask method with bilateral filter, which results in the improvement of classification accuracy. The texture features are extracted by using different values of range parameter and domain parameter and are fed as input to k-Nearest Neighbor (k-NN) classifier for classification. The new fusion model is tested with Brodatz, VisTex, STex and ALOT databases. The results of the proposed method are also compared with the conventional Laws' mask descriptor for all the aforementioned four datasets. The experimental results show that bilateral filter based Laws' mask feature extraction technique provides better classification accuracy for all the four databases for various combinations of bilateral filter range and domain parameters. (C) 2018 Production and hosting by Elsevier B.V. on behalf of Faculty of Computers and Information, Cairo University.

Keywords:

Bilateral filter

Laws mask descriptor

Feature extraction

Texture classification

AI Summary

Key information extracted from the uploaded paper, including a brief overview, abstract, background, key highlights, visual analysis, and future outlook.