The process of identifying regions with similar texture and separating regions with different texture is an essential step
towards identifying surfaces and objects in the image. This paper describes a method of processing discontinuities of the
gray level image. Moments are widely used in image analysis and pattern recognition. The Gaussian-Hermite moments,
as one kind of orthogonal moments, are proposed to estimate coherency of the image in this paper. The local image is
firstly converted into frequency domain from spatial domain, and then coherency is measured by matrix constructed of
Gaussian-Hermite moments of energy within frequency domain. Compared to other methods, such as coherency
estimation based on geometric moments, cross correlation, eigenstructure, semblance, or gradient vector field, the
experimental results show a good performance in feature representation, regions recognition and regions.