| K. N. Walker, T. F. Cootes and C. J. Taylor. Locating Salient Object Features. Proc. British Machine Vision Conference, pages 557-566, 1998. |
....We wish to evaluate the perceptual importance, or salience, of individual pixels within the source image. Reliably extracting such information from a general 2D image is a problem as yet unsolved by the vision community. However, we make progress by modifying a technique due to Walker et al.[7], who observe that such salient pixels are uncommon in an image. The basic technique is to model the statistical distribution of a set of measures associated with each pixel, and to isolate the outliers of this distribution. The pixels corresponding to these outliers are regarded as salient. To ....
C. J. T. K. N. Walker, T. F. Cootes. Locating salient object features. In BMVC Proceedings, volume 2, pages 557--567, 1998.
....literature, or elsewhere. However, we can make progress by choosing a definition of salience that is su#ciently general for our needs, and allow user interaction to provide power where it is needed. We locate salient features within a single image by modifying a technique due to Walker et al. [32], who observe that salient pixels are uncommon in an image. The basic technique is to model the statistical distribution of a set of measures associated with each pixel, and to isolate the outliers of this distribution. The pixels corresponding to these outliers are regarded as salient. To ....
K. N. Walker, T. F. Cootes, and C. J. Taylor, "Locating salient object features," in Proceedings BMVC, 1998, vol. 2, pp. 557-- 567.
....image feature that is present in only a single image or on a single object would allow to distinguish this image or object from all others. According to the definition, such a feature has maximum saliency. The term salient feature has previously been used by many other researchers, for example [2, 13, 16, 17], although definitions vary. Intuitively, saliency corresponds to the rarity of a feature. Following Walker et al. [17] we want to formalize this by defining the saliency over the probability density in feature space. In their work, they approximate the probability density by a mixture of ....
....all others. According to the definition, such a feature has maximum saliency. The term salient feature has previously been used by many other researchers, for example [2, 13, 16, 17] although definitions vary. Intuitively, saliency corresponds to the rarity of a feature. Following Walker et al. [17], we want to formalize this by defining the saliency over the probability density in feature space. In their work, they approximate the probability density by a mixture of Gaussians. This permits efficient use of the resulting saliency estimate, but the accuracy of the approximation is ....
K. N. Walker, T.F. Cootes, and Chris Taylor. Locating salient object features. In BMVC'98,
No context found.
K. N. Walker, T. F. Cootes and C. J. Taylor. Locating Salient Object Features. Proc. British Machine Vision Conference, pages 557-566, 1998.
....Also, the salient points selected from one face will vary due to expression, lighting, pose and identity changes, causing them to be confused with other points in new examples of a face. We have generalised this algorithm to be trained on sets of faces where a dense correspondence is known [17]. From this training set it is possible to analyse how features vary. Salient features can then be selected by choosing features which have a low probability of being misclassified with other features. We have shown that using such a training set can further improve the 15 selection of salient ....
K. N. Walker. Locating Salient Object Features. In British Machine Vision Conference 1998, pages 557--566, Southampton, UK, 1998. 19
No context found.
# K. N. Walker, T. F. Cootes and C. J. Taylor, Locating Salient Object Features, British Machine Vision Conference, 1998.
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