Texture Compression

Abstract

We characterize "visual textures" as realizations of a stationary, ergodic, Markovian process, and propose using its approximate minimal sufficient statistics for compressing texture images. We propose inference algorithms for estimating the "state" of such process and its "variability". These represent the encoding stage. We also propose a non-parametric sampling scheme for decoding, by synthesizing textures from their encoding. While these are not faithful reproductions of the original textures (so they would fail a comparison test based on PSNR), they capture the statistical properties of the underlying process, as we demonstrate empirically. We also quantify the tradeoff between fidelity (measured by a proxy of a perceptual score) and complexity.

Image patch analysis and clustering of sunspots: A dimensionality reduction approach

“…The Jtrend test shows that all of these relationships are significant except for the umbra PCA estimate. This is similar to [16] where the entropy of certain image textures is found to be generally increasing but nonmonotonically with scale. Figure 5 showsm of the single sunspot image at the different scales.…”

Steps Towards a Theory of Visual Information: Active Perception, Signal-to-Symbol Conversion and the Interplay Between Sensing and Control

“…A "texture" is a region of an image that exhibits some kind of spatial regularity. The following characterization of texture is adapted from [63], where the basic definitions of stationarity, ergodicity, and Markov sufficient statistics are described in some detail. The important issue for us is that, if a process is stationary and ergodic, a statistic φ can be predicted from a realization of the image in some region ω.…”

“…In the continuum, the problem above can be solved using variational optimization as in [168]. In the discrete, the reader can refer to [63] for a description of compression, synthesis, segmentation and rectification algorithms.…”

Fast and Scalable Earth Texture Synthesis using Spatially Assembled Generative Adversarial Neural Networks

“…The goal of texture synthesis is often to capture the spatial structural patterns and characteristics from given example images and create many images with similar statistical properties [1]. However, the earth geomaterials usually have complex textural properties that are often heterogeneous; random characteristics are formally defined as stationary, ergodic, and stochastic processes [1,2]. Sometimes the texture of earth geomaterials with complex topological geometry and compositions such as shale and carbonate rocks usually requires an expensive and time-consuming characterization process and the number of available samples for subsequent subsurface reservoir management design is usually small [3,4].…”