Endowing convolutional neural networks (CNNs) with the rotation-invariant capability is important for characterizing semantic contents of remote sensing (RS) images since they do not have typical orientations. Most existing deep methods learning CNN models are based on design proper or pooling layers, which aims at predicting correct category labels rotated RS equivalently. However, a few works...

Given a continuous dynamical system f on a compact metric space X and a continuous potential Φ : X → R, the generalized rotation set is the subset of R consisting of all integrals of Φ with respect to all invariant probability measures. The localized entropy at a point in the rotation set is defined as the supremum of the measuretheoretic entropies over all invariant measures whose integrals pr...

In this paper, we propose a new feature for texture representation that is based on pixel patterns and is independent of the variance of illumination and rotation. A gray scale image is transformed into a pattern map in which edges and lines used to characterize the texture information are classified by pattern matching. The Gabor filters can enhance edge features, however, are not effective in...

This paper introduces a combined scattering representation for texture classification, which is invariant to rotations and stable to deformations. A combined scattering is computed with two nested cascades of wavelet transforms and complex modulus, along spatial and rotation variables. Results are compared with state-of-the-art algorithms, with a nearest neighbor classifier.

A method for object recognition invariant under translation , rotation and scaling is addressed. The rst step of the method (preprocessing) takes into account the invariant properties of the normalized moment of inertia and a novel coding that extracts topological object characteristics. The second step (recognition) is achieved by using a Holographic Nearest Neighbor algorithm (HNN), where vec...

A new model for describing a three-dimensional (3D) trajectory is introduced in this article. The studied object is viewed as a linear combination of rotatable 3D patterns. The resulting model is now 3D rotation invariant (3DRI). Moreover, the temporal patterns are considered as shift-invariant. A novel 3DRI decomposition problem consists of estimating the active patterns, their coefficients, t...