Log-polar wavelet energy signatures for rotation and scale invariant texture classification Export

@article{citeulike:2905925, abstract = {Classification of texture images is important in image analysis and classification. This paper proposes an effective scheme for rotation and scale invariant texture classification using log-polar wavelet signatures. The rotation and scale invariant feature extraction for a given image involves applying a log-polar transform to eliminate the rotation and scale effects, but at same time produce a row shifted log-polar image, which is then passed to an adaptive row shift invariant wavelet packet transform to eliminate the row shift effects. So, the output wavelet coefficients are rotation and scale invariant. The adaptive row shift invariant wavelet packet transform is quite efficient with only O(n /spl middot/ log n) complexity. A feature vector of the most dominant log-polar wavelet energy signatures extracted from each subband of wavelet coefficients is constructed for rotation and scale invariant texture classification. In the experiments, we employed a Mahalanobis classifier to classify a set of 25 distinct natural textures selected from the Brodatz album. The experimental results, based on different testing data sets for images with different orientations and scales, show that the proposed classification scheme using log-polar wavelet signatures outperforms two other texture classification methods, its overall accuracy rate for joint rotation and scale invariance being 90.8 percent, demonstrating that the extracted energy signatures are effective rotation and scale invariant features. Concerning its robustness to noise, the classification scheme also performs better than the other methods.}, author = {Pun, Chi-Man and Lee, Moon-Chuen}, booktitle = {Pattern Analysis and Machine Intelligence, IEEE Transactions on}, citeulike-article-id = {2905925}, citeulike-linkout-0 = {http://dx.doi.org/10.1109/TPAMI.2003.1195993}, citeulike-linkout-1 = {http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=1195993}, comment = {*Very nice paper with straight-forward equations for implementation, images and examples, well explained* Summary: -objects: texture images -features: orientation and scale-invariant, derived from wavelet decompositions -classifier: Mahalanobis Overview: This is one of the most clear, straight-forward papers on this topic. It gives step-by-step explanations and implementations of the algorithm, nicely illustrated with flowcharts and images. The algorithm is based on QMF wavelet packet decomposition of images. The key point in the algorithm is the log-polar transformation of the image, which produces images who appear to be nearly orientation and scale-invariant. The resulting images also are shifted versions of each other, thus the wavelet decomposition that would also be shift invariant is needed. This problem is alleviated by having oct-tree decomposition (instead of standard quad-tree). The features used are energy signatures of subbands.}, doi = {10.1109/TPAMI.2003.1195993}, journal = {Pattern Analysis and Machine Intelligence, IEEE Transactions on}, keywords = {classification, orientation-invariant, scale-invariant, texture, wavelets}, number = {5}, pages = {590--603}, posted-at = {2008-06-18 19:09:38}, priority = {0}, title = {Log-polar wavelet energy signatures for rotation and scale invariant texture classification}, url = {http://dx.doi.org/10.1109/TPAMI.2003.1195993}, volume = {25}, year = {2003} }

TY - JOUR ID - citeulike:2905925 L3 - citeulike-article-id:2905925 N2 - Classification of texture images is important in image analysis and classification. This paper proposes an effective scheme for rotation and scale invariant texture classification using log-polar wavelet signatures. The rotation and scale invariant feature extraction for a given image involves applying a log-polar transform to eliminate the rotation and scale effects, but at same time produce a row shifted log-polar image, which is then passed to an adaptive row shift invariant wavelet packet transform to eliminate the row shift effects. So, the output wavelet coefficients are rotation and scale invariant. The adaptive row shift invariant wavelet packet transform is quite efficient with only O(n /spl middot/ log n) complexity. A feature vector of the most dominant log-polar wavelet energy signatures extracted from each subband of wavelet coefficients is constructed for rotation and scale invariant texture classification. In the experiments, we employed a Mahalanobis classifier to classify a set of 25 distinct natural textures selected from the Brodatz album. The experimental results, based on different testing data sets for images with different orientations and scales, show that the proposed classification scheme using log-polar wavelet signatures outperforms two other texture classification methods, its overall accuracy rate for joint rotation and scale invariance being 90.8 percent, demonstrating that the extracted energy signatures are effective rotation and scale invariant features. Concerning its robustness to noise, the classification scheme also performs better than the other methods. IS - 5 JF - Pattern Analysis and Machine Intelligence, IEEE Transactions on EP - 603 TI - Log-polar wavelet energy signatures for rotation and scale invariant texture classification VL - 25 T2 - Pattern Analysis and Machine Intelligence, IEEE Transactions on SP - 590 KW - classification KW - orientation-invariant KW - scale-invariant KW - texture KW - wavelets N1 - *Very nice paper with straight-forward equations for implementation, images and examples, well explained* Summary: -objects: texture images -features: orientation and scale-invariant, derived from wavelet decompositions -classifier: Mahalanobis Overview: This is one of the most clear, straight-forward papers on this topic. It gives step-by-step explanations and implementations of the algorithm, nicely illustrated with flowcharts and images. The algorithm is based on QMF wavelet packet decomposition of images. The key point in the algorithm is the log-polar transformation of the image, which produces images who appear to be nearly orientation and scale-invariant. The resulting images also are shifted versions of each other, thus the wavelet decomposition that would also be shift invariant is needed. This problem is alleviated by having oct-tree decomposition (instead of standard quad-tree). The features used are energy signatures of subbands. AU - Pun, Chi-Man AU - Lee, Moon-Chuen PY - 2003/// UR - http://dx.doi.org/10.1109/TPAMI.2003.1195993 ER -