Classification of Coral Reef Components Using Color and Texture Features

Abstract

This paper presents classification of coral reef benthic components that composed of live corals, dead corals, rubbles and sands. Since coral reef exist with different of shapes, colours and textures, the use of image processing technique provides advantages to estimate percentage cover of coral reef benthic components. Color and texture are used to extract features of coral reef benthic components. Hue Saturation Value (HSV) color model is utilized by calculating its color histogram to obtain color features. Meanwhile, the Local Binary Pattern (LBP) descriptor is used to extract texture features. The color and texture features are combined as the input into the Multilayer Perceptron Neural Network (MLPNN) classifier. The performances of the coral reef classification are evaluated based on color feature, texture feature or combination of both color and texture features. It is found out that the joining feature set of color and texture features provide the highest classification accuracy, i.e. 92.60% accuracy rate as compared to the use of individual feature such as color and texture features alone that achieved only 81.30% and 88.10% accuracy classification rate, respectively.