J.M. Francos, A.Z. Meiri, and B. Porat. A unified texture model based on a 2-D wold like decomposition. IEEE transactions on signal processing, 41:2665-- 2678, August 1993.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....samples at arbitrary angles and resolution. We have o#ered a solution, in principle: nearest neighbor HMM s are dense and can be estimated. Evidently, however, the approach is a long way from being practical. In any case, others have already made good progress we would cite [14] 25] 18] [21], 32] and [20] for some state of the art work on texture estimation and synthesis. 33 ....

J.M. Francos, A.Z. Meiri, and B. Porat. A unified texture model based on a 2-D Wold like decomposition. Technical Report, Department of Electrical Engineering, Technion-Israel Institute of Technology, Haifa 32000.

....imaged. 3 The model of texture proposed in the present work combines a structural element based on 2 Gamma D affine relationships between texture patches with a stochastic element designed to account for the unpredictable variations in the texture in a manner analogous to a Wold decomposition [15]. The next section contains an account of the new model. Examples are given of textures generated using the model and the issue of scale is addressed. A description is then given of an analysis by synthesis algorithm, which has been applied succesfully to the modelling of a number of textures ....

....affine one of figure 8(c) In effect, the texture is being modelled as the sum of a deterministic component , from the affine model, and a stochastic one. Such techniques have been used in audio signal analysis and synthesis for some time [33] and have recently been applied to texture modelling [15]. It is important not to push the analogy too far, however: the structural component defined by the affine model is not deterministic the randomness is in the model parameters, rather than the model input. A fundamental issue in this form of texture modelling is that of scale: is there an ....

J. M. Francos, A. Meiri, and B. Porat. A Unified Texture Model Based on a 2-D Wold-Like Decomposition. IEEE Trans. Signal Processing, 41:2665--2677, 1993.

....correlated for two reasons. First, the representation is highly overcomplete, so the coefficients lie within a linear subspace. More importantly, covariances of subband coefficients can arise from spectral peaks (i.e. periodicity) or ridges (i.e. globally oriented structure) in a texture [24]. In order to represent such spectral features, we should include the local autocorrelation of each subband as a texture descriptor. However, due to the overcompleteness of our linear representation, the spatial correlation of the subband responses are highly redundant, and thus unsuitable for a ....

....to be seen whether simple models of the joint statistics, such as that introduced in this paper, can provide a more efficient representation. Another interesting concept used in a number of recent models is to adapt the basis to the statistical properties of each individual texture example [6, 24, 48, 70, 42]. The model by Zhu et al. in particular, produces high quality synthesis examples using a surprisingly small set of filters. The drawback with this approach is the additional computational expense (often substantial) in the filter selection process. We envision a number of extensions of our ....

J M Francos, A Z Meiri, and B Porat. A unified texture model based on a 2-D Wold-like decomposition. IEEE Trans. Signal Proc., 41(8):2665--2678, 1993.

....in part by BT and by NEC. components: a purely indeterministic field, a generalizedevanescent field and a harmonic field [3] The background theory as well as some applications of this 2 D Wold like decomposition to spectral estimation and modeling of homogeneous textures can be found in [3] and [4]. It is necessary to note that the Wold based theory and applications presented to date in the literature assume the random field is stationary; this important assumption is usually violated in natural images. Inhomogeneities may even arise in homogeneous data simply by viewing it with perspective ....

....to the two less salient dimensions identified in the study of Rao and Lohse. The most general model for the purely indeterministic component is the moving average (MA) model. However, under certain assumptions, an auto regressive (AR) representation of this part of the random field exists [4]. Various implementations of auto regressive models have been used successfully for segmenting 4 8 textures in an image [7] In this work we use the simultaneous auto regressive (SAR) model of Mao and Jain [7] for the purely indeterministic component, as well as by itself for comparison to the ....

J. M. Francos, A. Zvi Meiri, and B. Porat. A unified texture model based on a 2-D Wold like decomposition. IEEE T. Sig. Proc., pages 2665--2678, August 1993.

....most commonly used low level features. A textured image region can often be regarded as a homogeneous (stationary) random field. The twodimensional (2 D) Wold like decomposition theory for homogeneous random fields has previously been introduced to texture analysis and synthesis in still images [1] and to periodic motion detection and segmentation in video [2] The 2 D Wold theory allows an image pattern to be decomposed into three mutually orthogonal components. The perceptual characteristics of these components can be described as periodicity , directionality , and randomness , ....

....and Fg ( j) correspond to spectral singularities supported by point like and line like regions, respectively. Examples of natural textures containing different prominent Wold components are shown in Figure 1. 3 Previous Work Two decomposition methods have been proposed in the literature [7] [1]. The first is a maximum likelihood direct parameter estimation procedure, which provides parametric descriptions of image Wold components. Its developers reported that the algorithm can be computationally expensive, especially when the number of spectral peaks is (a) b) c) Figure 1: Examples ....

[Article contains additional citation context not shown here]

J. M. Francos, A. Zvi Meiri, and B. Porat. A unified texture model based on a 2-D Wold-like decomposition. IEEE T. Sig. Proc., pages 2665--2678, August 1993.

....of these components can be described as periodicity , directionality , and randomness , agreeing closely with the most important dimensions of human texture perception. The 2 D Wold decomposition has been recently applied to spectral estimation and texture modeling by Francos et al. 2][3][4] In their work, it is assumed that the images are homogeneous random fields and the model designs are based on the actual image decomposition. Although their algorithms performed well on a few texture examples, they are not robust or computationally efficient enough to handle databases where ....

....regular homogeneous random field can be achieved by separating the singular and the absolutely continuous components of the SDF. This is known as Lebesgue decomposition [8] In order to apply the 2 D Wold theory to texture modeling, some approximations on the deterministic random fields were made [3]. A half plane deterministic field is approximated by a harmonic random field, which in the spectral domain appears as the 2 D Dirac ffi functions supported by discrete points. The SDF of an evanescent field is absolutely continuous in one dimension and singular in the orthogonal dimension, ....

[Article contains additional citation context not shown here]

J. M. Francos, A. Zvi Meiri, and B. Porat. A unified texture model based on a 2-D Wold like decomposition. IEEE T. Sig. Proc., pages 2665--2678, August 1993.

....matrices have also been used as texture features [8, 9] Although some of the features are orientation invariant, they lack the information needed to determine whether the change of scale and orientation is gradual or abrupt. Fourier spectrum can capture texture s frequency and orientation [4, 7]. Unfortunately, it is not localized in the spatial domain. It is impossible to extract a textured region s Fourier spectrum unless the region has already been segmented. Gabor filters have the advantage of being optimally localized simultaneously in the spatial and the spatial frequency domains. ....

J. M. Francos, A. Zvi Meiri, and B. Porat. A unified texture model based on a 2-D Wold like decomposition. IEEE Transactions on Signal Processing, pages 2665-- 2678, Aug. 1993.

....properties of the three components can be described as periodicity , directionality , and randomness , agreeing closely to the important dimensions of human texture perception. The 2 D Wold decomposition have been recently applied to spectral estimation and texture modeling by Francos et al. 2][3]. In this paper, we present a new Wold based texture model and its application to image retrieval in large texture databases. The Wold based texture modeling presented to date in the literature assumes that the images are stationary random fields, and, therefore, the model implementations are not ....

....facilitating pattern comparison in real time. The Wold model implementations in the literature were not designed to meet these requirements; therefore, a new implementation is necessary. The Wold model implementations reported to date take one of two approaches. One is Lebesgue decomposition [2] [3], and the other is direct maximum likelihood (ML) parameter estimation [10] Compared to the ML method, the algorithms based on Lebesgue decomposition are computationally more efficient. Also, Fourier spectral analysis has the advantage of being shift invariant; humans, too, are relatively ....

J. M. Francos, A. Zvi Meiri, and B. Porat. A unified texture model based on a 2-D Wold like decomposition. IEEE T. Sig. Proc., pages 2665--2678, August 1993.

No context found.

J. M. Francos, A. Z. Meiri, and B. Porat, A unified texture model based on a 2-D Wold-like decomposition, IEEE Trans. Signal Proc. 41 (1993), 2665-2678.

.... random fields analysis, a similar type of support was used by Whittle [W] as well as by Ekstrom and Woods [EW] to develop the concept of 2 D spectral factorization; by Marzetta [M] to describe a theoretical solution of the 2 D normal equations by a 2 D Levinson type algorithm; and in [FMP2], FNW] to implement an analysis synthesis procedure for homogeneous texture fields. Helson and Lowdenslager [HL2] proved some of the results contained in sections 3 and 4 for the case of homogeneous random fields using the character group approach. However, frequency domain analysis is applicable ....

J. M. Francos, A. Z. Meiri, and B. Porat, A unified texture model based on a 2-D Wold-like decomposition, IEEE Trans. Signal Proc. 41 (1993), 2665-2678.

No context found.

J.M. Francos, A.Z. Meiri, and B. Porat. A unified texture model based on a 2-D wold like decomposition. IEEE transactions on signal processing, 41:2665-- 2678, August 1993.

No context found.

Joseph M. Francos, A. Zvi Meiri, Boaz Porat, A Unified Texture Model Based on a 2-D Wold-Like Decomposition, IEEE Trans. on Signal Processing, vol 41(8), pp. 2665-2678, 1993.

No context found.

J. M. Francos, A. Z. Meiri, and B. Porat. A unified texture model based on a 2-d wold like decomposition. IEEE Trans. SP, 41:2665--2678, 1993.

No context found.

J. M. Francos, A. Zvi Meiri, and B. Porat. A unified texture model based on a 2-d wold-like decomposition. IEEE Transactions on Signal Processing, pages 2665-- 2678, August 1993.

No context found.

J M Francos, A Z Meiri, and B Porat. A unified texture model based on a 2-D Wold-like decomposition. IEEE Trans. Signal Proc., 41(8):2665--2678, 1993.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC