Because object geometry varies at many scales, it is often convenient to distinguish shape from texture. While shape is a deterministic macroscopic description, texture is a finer scale geometric description with some repetitive or random component. The distinction between texture and shape is important when developing object recognition systems. Acquiring fine scale geometry is difficult due t...

Here we demonstrate that the feature space of random shallow convolutional neural networks (CNNs) can serve as a surprisingly good model of natural textures. Patches from the same texture are consistently classified as being more similar then patches from different textures. Samples synthesized from the model capture spatial correlations on scales much larger then the receptive field size, and ...

The statistics of gray level differences have been successfully used in a number of texture analysis studies. In this paper we propose to use signed gray level differences and their multidimensional distributions for texture description. The present approach has important advantages compared to earlier related approaches based on gray level cooccurrence matrices or histograms of absolute gray l...

A distribution-based classification approach and a set of recently developed texture measures are applied to rotation-invariant texture classification. The performance is compared to that obtained with the well-known circular-symmetric autoregressive random field (CSAR) model approach. A difficult classification problem of 15 different Brodatz textures and seven rotation angles is used in exper...

This paper empirically compares nine image dissimilarity measures that are based on distributions of color and texture features summarizing over 1,000 CPU hours of computational experiments. Ground truth is collected via a novel random sampling scheme for color, and via an image partitioning method for texture. Quantitative performance evaluations are given for classification, image retrieval, ...

Many components and devices in electronic systems are fabricated from materials that have a preferred crystallographic orientation or texture. The applications in which the texture of the material plays a key role in determining the properties and performance are broad: Al and Cu interconnects in integrated circuits, complex oxides in random access memory devices, and metallic alloys in magneti...

Various methods have been developed for texture segmentation. Since all of them have their merits and drawbacks, the choice of the most suitable method becomes a nontrivial task. We present a comparative study of texture segmentation methods based on four frequently used texture feature extraction techniques: gray level co-occurrence matrices, Gaussian Markov random fields, Gabor filtering, and...

This paper describes basic steps of a new technique, called bunch sampling, that allows to realistically synthesise spatially homogeneous textures. A geometric shape of the bunch (acting as a texel) and spatial placement grid governing relative positions of the bunches are estimated from the training texture by using a generic Gibbs random field texture model with multiple pairwise pixel intera...

Texture models are widely in use for image content description. In remote-sensing images textures occur at very different scales, requiring the application of multiple texture models. We present an algorithm based on a multi-scale random field model to detect the characteristic scales at which textures are present, so that texture models can be applied to a few selected scales only. The algorit...

An efficient and robust type of unsupervised multispectral texture segmentation method is presented. Single decorrelated monospectral texture factors are assumed to be represented by a set of local Gaussian Markov random field (GMRF) models evaluated for each pixel centered image window and for each spectral band. The segmentation algorithm based on the underlying Gaussian mixture (GM) model op...