Evaluation of various filter kernels for extraction of linear features from satellite imagery

Abstract

Edge-enhancement filters, among others, are tools for image enhancements, which provide the user with more options for delineating linear features. The aim of this paper is to study various filter kernel combinations to ascertain their value and usability for extraction of linear features in particular. Twelve different filter kernels of 3x3 matrices were studied, which included six of zero-sum kernel type and six of high-pass filter type. Eleven filters showed satisfactory result for edge-enhancement. The use of filters for extraction of linear features from satellite imagery was found to be very suitable, and it shows that other filters may be defined according to the purpose of the user.