A. Laine and J. Fan, "Frame Representation for Texture Segmentation," IEEE Trans. Image Processing, vol. 5, no. 5, pp. 771-779, 1996.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....is selected for use in the composite representation [2] The rationale is that large absolute coefficients often correspond to salient features in the images. However, obviously this simple selection rule does not always work. In this paper, we use the discrete wavelet frame transform (DWFT) [1] (also called the stationary wavelet transform [3] to perform multiresolution decomposition, and then use the support vector machine (SVM) 4] to fuse the wavelet frame coefficients. DWFT is closely related to DWT but avoids down sampling by using an overcomplete wavelet decomposition. Its ....

A. Laine and J. Fan. Frame representations for texture segmentation. IEEE Transactions on Image Processing, 5(5):771--780, 1996.

....Mirror Filters [17] Note that these requirements are not essential for segmentation and classification tasks. Rather we require that the representation is stable and permits the extraction of features to adequately characterize textural properties; therefore, one often employs redundant frames [36]. Moreover, the decomposition in the twodimensional space is usually implemented by applying the one dimensional filters separably. One then is confronted with the problem of angular selectivity. Separable filters are strongly oriented in the horizontal and vertical directions [17] 37] For its ....

A. Laine and J. Fan. Frame representation for texture segmentation. IEEE Trans. Im. Proc, 13(4):1--8, 1995.

....most of which obey a two stage scheme [1, 2, 3, 4] 1. In the modeling stage characteristic features are extracted from the textured input image, which range from spatial frequencies [5, 2, 4] MRF models [6, 7, 3, 8, 9] co occurrence matrices [10] to wavelet coecients [11] wave packets [12] and fractal indices [13] 2. In the optimization stage features are grouped into homogeneous segments by minimizing an appropriate quality measure. This is most often achieved by a few types of clustering cost functions [1, 2, 7, 3, 8, 9, 13, 12, 4, 14] Occasionally, additional a priori ....

....9] co occurrence matrices [10] to wavelet coecients [11] wave packets [12] and fractal indices [13] 2. In the optimization stage features are grouped into homogeneous segments by minimizing an appropriate quality measure. This is most often achieved by a few types of clustering cost functions [1, 2, 7, 3, 8, 9, 13, 12, 4, 14]. Occasionally, additional a priori information is used in spatial interaction models [1, 15] The strict separation of the modeling and the optimization stage is important since it enables an evaluation of the quality and speed of optimization procedures independently from the feature extraction ....

[Article contains additional citation context not shown here]

A. Laine and J. Fan, \Frame representations for texture segmentation," IEEE Transactions on Image Processing, vol. 5, no. 5, pp. 771-779, 1996.

.... application of texture analysis algorithms for characterizing intramuscular fat percentage in ribeye muscle of the cattle has been described [6] Among the many texture analysis algorithms, wavelet based methods have also shown significant potential for texture classification and segmentation [8] [9]. We present the use of wavelet transform for tissue texture analysis to evaluate tendon injuries, in this paper. 2 TENDON IMAGE ACQUISITION Ultrasound scans were obtained from horses that injured their tendons while racing in 1996. Images were collected from 24 horses with SDFT injury. Of ....

....Image Analysis A texture segmentation method using discrete wavelet packet frame (DWPF) was applied to the cross sectional image to differentiate the damaged lesion and to quantify the percentage of the damaged SDFT. The DWPF is an overcomplete version of wavelet packet decomposition [9], where downsampling is not involved. Laine and Fan [9] show some promising results for texture image segmentation using DWPF. Figure 3 shows the structure for DWPF (decomposition stages 0, 1 only) The wavelet filters H (w) and G l (w) at level l are generated using the equations in the figure. ....

[Article contains additional citation context not shown here]

Laine, A. and J. Fan, Frame representations for texture segmentation, IEEE Trans. on Image Processing, Vol. 5, No. 5, pp.771-780, 1996.

....the past decades, most of which obey a two stage scheme: 1. A modeling stage: characteristic features are extracted from the textured input image, which range from spatial frequencies [1, 2, 3] MRF models [4, 5, 6, 7, 8] co occurrence matrices [9] to wavelet coefficients [10] wave packets [11] and fractal indices [12] 2. In the optimization stage features are grouped into homogeneous segments by minimizing an appropriate quality measure. This is most widely achieved by a few types of clustering cost functions [2, 5, 6, 7, 8, 12, 11, 3] Occasionally additional a priori information is ....

....co occurrence matrices [9] to wavelet coefficients [10] wave packets [11] and fractal indices [12] 2. In the optimization stage features are grouped into homogeneous segments by minimizing an appropriate quality measure. This is most widely achieved by a few types of clustering cost functions [2, 5, 6, 7, 8, 12, 11, 3]. Occasionally additional a priori information is used in spatial interaction models [9, 13] The strict separation of the modeling and the optimization stage is important in that it enables us to evaluate the quality and speed of optimization procedures independently from the feature extraction ....

[Article contains additional citation context not shown here]

A. Laine and J. Fan, "Frame representations for texture segmentation," IEEE Transactions on Image Processing, vol. 5, no. 5, pp. 771--779, 1996.

....most of which obey a two stage scheme [1, 2, 3, 4] 1. In the modeling stage characteristic features are extracted from the textured input image, which range from spatial frequencies [5, 2, 4] MRF models [6, 7, 3, 8, 9] co occurrence matrices [10] to wavelet coefficients [11] wave packets [12] and fractal indices [13] 2. In the optimization stage features are grouped into homogeneous segments by minimizing an appropriate quality measure. This is most often achieved by a few types of clustering cost functions [1, 2, 7, 3, 8, 9, 13, 12, 4, 14] Occasionally, additional a priori ....

....co occurrence matrices [10] to wavelet coefficients [11] wave packets [12] and fractal indices [13] 2. In the optimization stage features are grouped into homogeneous segments by minimizing an appropriate quality measure. This is most often achieved by a few types of clustering cost functions [1, 2, 7, 3, 8, 9, 13, 12, 4, 14]. Occasionally, additional a priori information is used in spatial interaction models [1, 15] The strict separation of the modeling and the optimization stage is important since it enables an evaluation of the quality and speed of optimization procedures independently from the feature extraction ....

[Article contains additional citation context not shown here]

A. Laine and J. Fan, "Frame representations for texture segmentation," IEEE Transactions on Image Processing, vol. 5, no. 5, pp. 771--779, 1996. J. Puzicha, J.M. Buhmann: Real--Time Texture Segmentation 24

....unsupervised. As mentioned before, the local feature calculation is translation variant. To overcome this subsampling is often omitted when using the DWT in segmentation tasks. One then obtains an overcomplete representation in which the redundancy improves robustness. Unser [18] and Laine and Fan [30] report on this. In general, omission of subsampling will improve results, but increase computation time. In [31] extensive studies on the effects of this are reported. In segmentation it can be advantageous to include spatial information in the feature set. A simple way to do this is to add the ....

A. Laine and J. Fan, "Frame representations for texture segmentation," IEEE Transactions on Image Processing, vol. 5, no. 5, pp. 771--779, 1996.

....in which the redundancy improves robustness. The decomposition also becomes translation invariant and the correspondence between scales is trivial. In wavelet packet decompositions, the non subsampled scheme is referred to as wavelet frame decomposition . Unser [33] and Laine and Fan [20] report on this. In general, omission of subsampling will improve results, but increase computation time. Randen and Husoy [27] have made extensive studies on the effects of this. III. Other Issues A. Which wavelet to choose Several studies have tried to answer this important question. There ....

A. Laine and J. Fan. Frame representations for texture segmentation. IEEE Transactions on Image Processing, 5(5):771--779, 1996.

....[33, 34] co occurrence matrices [16] to fractal indices [6] 2. In the clustering stage features are grouped into homogeneous segments, where homogeneity of features has to be formalized by a mathematical notion of similarity. Most widely, features are interpreted as vectors in a Euclidean space [25, 33, 34, 65, 40, 6, 31] and a segmentation is obtained by minimizing the K means criterion, which sums over the square distances between feature vectors and their assigned, group specific prototype feature vectors. K means clustering can be understood as a statistical mixture model with isotropic Gaussian class ....

A. Laine and J. Fan. Frame representations for texture segmentation. IEEE Transactions on Image Processing, 5(5):771--779, 1996.

No context found.

A. Laine and J. Fan, "Frame Representation for Texture Segmentation," IEEE Trans. Image Processing, vol. 5, no. 5, pp. 771-779, 1996.

No context found.

A. Laine and J. Fan, "Frame Representations for Texture Segmentation", IEEE Transactions on Image Processing, Vol.5 No. 5, MAY. 1996, pp.771-780.

No context found.

A. Laine and J. Fan. Frame representation for texture segmentation. IEEE Trans. Image Processing, 5:771--780, May 1996.

No context found.

#20# A. Laine and J. Fan. Frame representations for texture segmentation. IEEE Transactions on Image Processing,

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