Image texture synthesis-by-analysis using moving-average models

Abstract

Texture synthesis is a necessary component of realistic scene generation. In particular, it is necessary for the simulation of image backgrounds for the testing of automatic target recognizers. We present a synthesis-by-analysis model for texture replication or simulation. This model can closely replicate a given textured image or produce another image which, although distinctly di erent from the original, has the same general visual characteristics and the same rst and second-order gray-level statistics as the original image. In e ect, such a synthetic image looks like a continuation of the original scene; as if another picture of the scene were taken adjacent to the original. The texture synthesis algorithm proposed herein contains three distinct components: A moving-average (MA) lter, a lter excitation function, and a gray-level histogram. The analysis portion of texture synthesis algorithm derives the three from a given image. The synthesis portion convolves the MA lter kernel with the excitation function, adds noise, and modi es the histogram of the result. The advantages of this texture model over others include conceptually and computationally simple and robust parameter estimation, inherent stability, parsimony in the number of parameters, and synthesis through convolution. We herein (1) describe a procedure for deriving the correct MA kernel using a signal enhancement algorithm; (2) demonstrate the e ectiveness of the model by using it to mimic several diverse textured images; (3) discuss its applicability to the problem of infrared background simulation; and (4) include detailed algorithms for the model's implementation.

Robust Rotation-Invariant Texture Classification Using a Model Based Approach

“…In essence, the model replaces usual independent and identically distributed (i.i.d.) sources employed in linear texture models [13]- [15] with a non i.i.d. binary source.…”

“…Moreover, in the last decade, wavelet theory has been widely used for texture classification purposes [8]- [10]. Several stochastic models have also been proposed for texture modeling and classification; they include Gaussian Markov random fields models [11], [12], moving average (MA), autoregressive (AR), and autoregressive moving average (ARMA) models [13]- [15].…”

A multiresolution approach for texture synthesis using the circular harmonic functions

“…excitation used in classical approaches (see, for example, [4] and [7]) to feed a linear system is replaced by a more structured field tailored to retain the morphological behavior of the given prototype. The choice of a binary excitation presents the advantage that, under weak symmetry conditions, the whole of its second-order distributions is set by its auto correlation function (acf).…”

“…The quality of the obtained synthetic textures is related to the actual compliance of the patterns with a convolutional model. However, these models are often unable to preserve the underlying structure of the given prototype so that, in order to achieve a better texture quality, a human supervised step is often required for the definition of a more structured excitation [7].…”

Nonlinear model selection for PARMA processes using RJMCMC

“…ARMA models have been used in diverse areas of applications such as speech [2], [3], seismology [4], video [5], image [6], etc. Particularly, they have been applied in energy and meteorological prediction studies of solar radiation [7], [8], electricity demand [9], [10] and wind speed [11], [12].…”