Texture synthesis‐by‐analysis method based on a multiscale early‐vision model

Abstract

A new texture synthesis-by-analysis method, applying a visually based approach that has some important advantages over more traditional texture modeling and synthesis techniques is introduced. The basis of the method is to encode the textural information by sampling both the power spectrum and the histogram of homogeneously textured images. The spectrum is sampled in a log-polar grid using a pyramid Gabor scheme. The input image is split into a set of 16 Gabor channels (using four spatial frequency levels and four orientations), plus a lowpass residual (LPR). The energy and equivalent bandwidths of each channel, as well as the LPR power spectrum and the histogram, are measured and the latter two are compressed. The synthesis process consists of generating 16 Gabor filtered independent noise signals with spectral centers equal to those of the Gabor filters, whose energy and equivalent bandwidths are calculated to reproduce the measured values. These bandpass signals are mixed into a single image, whose LPR power spectrum and histogram are modified to match the original features. Despite the coarse sampling scheme used, very good results have been achieved with nonstructured textures as well as with some quasiperiodic textures. Besides being applicable to a wide range of textures, the method is robust (stable, fully automatic, linear, and with a fixed code length) and compact (it uses only 69 parameters).

Machine Vision based Detection of Defectsin Textile (Fabric) Material

“…There are three principal approaches in image processing are statistical, structural and spectral [6]. Statistical approaches yield characterization of textures as smooth, grainy and so on.…”

Invariant texture segmentation via circular Gabor filters

“…A great deal of work (including physiological research and theoretical algorithms) has been done on this important issue [7][9][17] [18]. Many methods have been proposed to reduce the degrees of freedom of Gabor filters based on neurophysiological findings [17]. One important finding is that the spatial-frequency bandwidth of the response of the human visual cortical cells ranges from 0.5 to 2.…”

Self-Invertible 2D Log-Gabor Wavelets

“…We pay special attention to the different aspects of the multiresolution design and we propose a series of solutions for improving the efficiency of the Gabor wavelets scheme. In comparison with previous implementations of Gabor multiresolution (Gross and Koch, 1995;Nestares et al, 1998;Ro et al, 2001;Portilla et al, 1996), the most important novelties of the present work consists in incorporating complex-valued oriented high-pass filters, and achieving the exact reconstruction and self-invertibility properties.…”

“…Gabor multiresolutions have been successfully used for image analysis and applications where exact reconstruction is not required, such as texture analysis (Clausi and Jernigan, 2000;Ro et al, 2001), texture synthesis (Portilla et al, 1996), edge/contour extraction (Heitger et al, 1998;Kovesi, 2003;Grigorescu et al, 2003), or object recognition (Pötzsch et al, 1996;Krüger, 2001). And, even without exact reconstruction they have been shown useful for image restoration applications (Cristóbal and Navarro, 1994;Kovesi, 1999;Mingolla et al, 1999;Christiansen, 2002).…”

“…The low-pass filters could be defined simply as a Gaussian low-pass filter as in Portilla et al (1996). Nevertheless for a better filling-in of the low-pass residual frequencies not covered by the log-Gabor filters, it is defined in a Fig.…”