| Tao-I. Hsu. Texture Analysis and Synthesis using the Multiresolution Fourier Transform. PhD thesis, University of Warwick, UK, 1994. |
....While at the highest level, each of the N Theta N blocks covers only one point in spatial domain, so the MFT is just the original image itself. The other levels in between are STFTs with windowing functions of different size. MFT has following merits which are important to image analysis [3] [7]: i)Space Linearity: MFT is a hierarchical structure of STFT s, and because STFT is linear, therefore MFT inherits the linear property. Ensuring linearity, the response of the filtering operation on the signal and addition can be predicted. ii)Locality: each level of MFT performs local ....
....with artifact of a linear feature. In this work, artifacts are not predicted and not tackled, and this affects the accuracy of orientation estimation. 3. 3 Estimation of centroid vectors Hsu has used a pair of centroid vectors to represent a local spectrum for texture analysis and synthesis [7]. Fourier spectrum is used to implement his algorithm. Similar approach is used in this work. But after some experiment, power spectrum , in stead of Fourier spectrum, is used to implement the following algorithm in this work in order to get more accurate feature orientation. Details about the ....
T. I. Hsu. Texture Analysis and Synthesis using Multiresolution Fourier Transform. PhD thesis, Department of Computer Science, The University of Warwick, UK, 1994.
....image. It has been mentioned that finding affine motion parameters can be a time consuming job when done by block matching. Therefore, in this work the motion parameters are calculated in a more efficient way within the framework of the MFT. The estimation is based on a method developed by Hsu [10] in the context of texture analysis and synthesis. 1.2 Feature Extraction One of the novelties of the proposed video compression method is that it makes use of linear features. So the question arises of how best to extract these features from the images in the sequence. The simplest kind of ....
T. Hsu. Texture analysis and synthesis using the Multiresolution Fourier Transform. PhD thesis, Department of Computer Science, The University of Warwick, UK, September 1994.
....one of the measures derived from the GLCM, the best match can be found between two images. The Multiresolution Fourier Transform(MFT) 11] is a linear transformation providing local frequency estimation over multiple spatial resolutions which could help analysing the motion between two images[7]. This work, therefore, will give an assessment of computation cost to the image matching by comparing GLCM and cross correlation method. There are four major steps in this work. Firstly, a Laplacian pyramid of a pair of images is constructed by using a low pass lter. Secondly, the ....
T.I. Hsu. Texture Analysis and Synthesis Using the Multiresolution Fourier Transform. PhD thesis, Department of Computer Science, The University of Warwick, UK, 1994.
.... Gamma (l) is a window function. Therefore each level of MFT resembles a 2 D Short Time Fourier Transform (STFT) x( 0 (l) is the original N Theta N image sampled at the point 0 at level l in the spatial domain. The MFT has following important properties for image analysis[1][7][8] 1) Linearity: MFT is a hierarchical set of STFT s, and the STFT is linear by definition. Hence, the MFT is linear. 2) Locality: a level of the MFT is a combined spatial and spatial frequency representation of the image. By choosing analysis vectors that are opti3 mally localised, each ....
T. I. Hsu. Texture Analysis and Synthesis using Multiresolution Fourier Transform. PhD thesis, Department of Computer Science, The University of Warwick, UK, 1994.
....are associated with regions defined on a quadtree tessalation of the image plane. The estimator consists of two components: 1. A local motion estimation scheme, in which the six parameters of the affine motion model are estimated using a correlation technique implemented via the frequency domain [15, 16]. 2. An efficient multiresolution search procedure to determine the final motion field in which motion estimates obtained at larger regions are used to direct the estimation of those at smaller region sizes [13, 14] The result is a motion field associated with each frame similar to that shown in ....
....scheme to determine optical flow. An affine model is then fitted to the found optical flow by means of some least squares fitting procedure. The affine model and estimation scheme used here is based on the frequency domain approach used by Hsu et al. for texture representation and analysis [15, 16]. The method utilises the fact that the magnitude of the Fourier transform is invariant to translation and equivariant to linear transformation [18] This enables the estimation of a full affine transformation to be separated into the finding of the linear component and the subsequent finding of ....
[Article contains additional citation context not shown here]
T. I. Hsu, Texture Analysis and Synthesis using the Multiresolution Fourier Transform. PhD thesis, University of Warwick, UK, September 1994.
....of the approach for a range of images and transformations. 3. 1 Local Affine Motion Estimation The procedure used for estimating the affine transformation relating two local regions is adapted from a frequency domain approach originally described by Hsu, Calway and Wilson for texture synthesis [65, 64, 63, 80]. It utilises the affine theorem for the 2 D Fourier transform [21] Given two continuous images, x 1 ( and x 2 ( related by an affine transformation such that x 1 ( x 2 (A b) 3.1) where is the spatial coordinate vector, A is an invertible 2 Theta 2 matrix representing the linear ....
T.-I. Hsu. Texture analysis and synthesis using the multiresolution Fourier transform. PhD thesis, University of Warwick, UK, September 1994.
No context found.
Tao-I. Hsu. Texture Analysis and Synthesis using the Multiresolution Fourier Transform. PhD thesis, University of Warwick, UK, 1994.
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