Texture classification of segmented regions of FLIR images using neural networks

Published 17 December 2002

J.F. Haddon, J.F. Boyce

DOI10.1109/icip.1994.413653

Citations8

TL;DR

The major regions in a sequence of FLIR images, segmented using co-occurrence techniques, are subjected to texture classification, which helps to train a range of multilayer perceptron neural networks using autonomously segmented regions in every fifth image.

Abstract

The major regions in a sequence of FLIR images, segmented using co-occurrence techniques, are subjected to texture classification. Texture can be thought of as a measure of the edginess about a pixel and can thus be described by edge co-occurrence matrices. These are temporally smoothed to ensure consistency in segmentation between the images of a sequence. The matrix of each segmented region is decomposed using 2-dimensional orthogonal Hermite functions. The coefficients of the decomposition are regarded as a low order feature vector which is characteristic of the region texture. Principal component analysis is used to determine the optimum combination of coefficients. In an application for use with infrared imagery these are then used to train a range of multilayer perceptron neural networks using autonomously segmented regions in every fifth image. The best network is then used to classify the remainder of the 300 images.>

Keywords

ChemistryEngineering

IEEE Transactions on Pattern Analysis and Machine IntelligenceA Computational Approach to Edge Detection

29,107 Citations1986John Canny

There is a natural uncertainty principle between detection and localization performance, which are the two main goals, and with this principle a single operator shape is derived which is optimal at any scale.

IEEE Transactions on Systems Man and CyberneticsTextural Features for Image Classification

22,499 Citations1973Robert M. Haralick, K. Shanmugam +1 more

These results indicate that the easily computable textural features based on gray-tone spatial dependancies probably have a general applicability for a wide variety of image-classification applications.

Computer Vision Graphics and Image ProcessingImage segmentation techniques

2,226 Citations1985Robert M. Haralick, Linda G. Shapiro

Elsevier eBooksStatistical Pattern Recognition

500 Citations1990E.R. Davies

This chapter discusses statistical pattern recognition and pruning the training set by eliminating patterns that are not near the boundaries of class regions in feature space, as such patterns do not materially help in reducing the misclassification rate.

IEEE Transactions on Pattern Analysis and Machine IntelligenceImage segmentation by unifying region and boundary information

209 Citations1990J.F. Haddon, J.F. Boyce

A two-stage method of image segmentation based on gray level cooccurrence matrices that robustly segments an image into homogeneous areas and generates an edge map is described and extends easily to general edge operators.

Image and Vision ComputingEdge detection and motion detection

117 Citations1986Libor Spacek

An integrated theory of edge detection, curvature measurement and motion detection during the earliest stages of visual processing is presented.

Electronics & Communications Engineering JournalCo-occurrence matrices for image analysis

59 Citations1993J.F. Haddon, J.F. Boyce

The authors present a range of techniques for image segmentation and edge detection based on co-occurrence matrices which can be integrated easily with a variety of postprocessing techniques such as hysteresis and relaxation labelling for enhanced performance.

Texture segmentation and region classification by orthogonal decomposition of co-occurrence matrices

12 Citations2003J.F. Haddon, J.F. Boyce

Techniques are presented for determining the periodicity and scale of texture and this information is used to define an adaptive transform based on edge co-occurrence matrices for segmenting textured regions.

Neural Networks for the Texture Classification of Segmented Regions of Forward Looking Infrared Images

1 Citations1993J.F. Haddon, J.F. Boyce +2 more

Texture can be interpreted as a measure of the 'edginess' about a pixel and can thus be described by edge co-occurrence matrices, the coefficients of which provide a low order feature vector which is characteristic of the texture.