Computer Vision, Graphics, and Image Processing
The Design and Use of Steerable Filters
IEEE Transactions on Pattern Analysis and Machine Intelligence
Multiresolution circular harmonic decomposition
IEEE Transactions on Signal Processing
Optimal detection and estimation of straight patterns
IEEE Transactions on Image Processing
Local orientation analysis in images by means of the Hermite transform
IEEE Transactions on Image Processing
Image denoising using scale mixtures of Gaussians in the wavelet domain
IEEE Transactions on Image Processing
The multiscale Hermite transform for local orientation analysis
IEEE Transactions on Image Processing
| Hi-index | 0.00 |
Identifying and estimating the orientation of monodimensional (linear) patterns found into images is an important task for pattern recognition purposes (e.g., in SAR images) and can greatly improve the efficacy of image restoration, sharpening and de-noising procedures. Existing approaches to linear pattern orientation estimation are based either on pyramid filter banks, steered to a small set of discrete orientations, or on parametric approaches based on the tensor gradient. In the present work, using a local tomography paradigm, the complementary properties of Hermite and Gauss-Laguerre image expansions are exploited for accurately estimating the orientation angles of linear patterns by rooting a polynomial, built with transform coefficients at each analysis point of the image. In particular, as in direction finding with sensor arrays (e.g., ROOT-MUSIC, MODE and ESPRIT), rooting allows a fast and accurate orientation estimation on a continuous set of angles. It is shown that the feasibility of this scheme is based on the simple link existing between Hermite and Gauss-Laguerre coefficients.