Pixel-Based Texture Classification by Integration of Multiple Texture Feature Evaluation Windows

Abstract

A wide variety of texture feature extraction methods have been proposed for texture based image classification and segmentation. These methods are typically evaluated over windows of the same size, the latter being usually chosen for each particular method on an experimental basis. This paper shows that pixel-based texture classification can be significantly improved by evaluating a given texture method over multiple windows of different size and then by integrating the results through a classical Bayesian scheme. The proposed technique has been applied to well-known families of texture methods that are frequently utilized for feature extraction from textured images. Experiments show that the integration of multisized windows yields lower classification errors than when optimal single-sized windows are considered.

Automatic texture feature selection for image pixel classification

“…[1,2]. Then, the application of the proposed selection scheme to the problem of texture feature selection is illustrated by adapting it to the proposed classification algorithm.…”

“…Moreover, the utilization of the selection methodology presented in this paper along with the classifier introduced in Refs. [1,2] produces better classification results than extensively used texture classifiers based on texture methods belonging to the same family.…”

“…The proposed selection scheme is applied to the problem of texture feature selection and incorporated to the texture classifier previously introduced in Refs. [1,2] as a preprocessing, off-line stage.…”

“…Section 4 details the application of the proposed methodology to the problem of texture feature selection and integrates this selector with the texture classification algorithm introduced in Refs. [1,2]. Section 5 shows experiments where the selection technique proposed in this paper is applied and evaluates the classification results and the achieved efficiency.…”

“…In its most general form, the problem consists of identifying the texture pattern to which every image pixel belongs [1][2][3]. This problem differs from texture segmentation, which aims at finding regions of uniform texture within the image without identifying them.…”