Edge detection Using Combination of Decomposition Filter and Simplified
Hyperbolic Tangent Filter
Abstract
This paper presents a new combination of Decomposition Filters (DF) and
Hyperbolic Tangent (HBT) based simplified directional operators for edge
detection (ED). Conventional separable algorithms are limited in
capturing the geometric features and non-separable filters eliminate
these restrictions. The high-frequency band of the DF of proposed method
achieves the edge information. A set of edge operators in the form of
simple non-separable patterns for different scales, and orientations are
applied on the edge information to capture linear edge structures and
directions. While testing with ten natural images, this algorithm has
improved performances in terms of reduced Mean Squared Error (MSE) in
the reconstruction estimation measure. The performance measure compares
the Simplified Gabor Wavelet (SGW) based ED.