Linear feature enhancing by Hyperbolic-Gaussian filtering

Abstract

Abstract Abstract. A new methodology for the linear feature enhancing of an image by filtering in Fourier space is presented. As a first step, the Fourier spectrum of the image is used to construct its rose diagram. A new filter is then applied in enhancement directions which are chosen directly from the rose diagram. In Fourier space, the proposed filter's transfer function is a surface whose section perpendicular to the enhanced direction is a Gaussian curve. The construction of this surface is made in such a way that filtering in any direction is allowed. A sequence of analysis for the construction of an enhanced image using this filter is proposed. As an optional step, after constructing the enhanced image a binariza-tion can be made to facilitate the linear feature map extraction. The proposed binarization corresponds to the application of a local threshold technique. A pseudocode summary of each of the steps is included. A comparison with a simple directional filter is made.