E. Borenstein and S. Ullman. Class-specific, top-down segmentation. In Proc. 7th European Conf. on Computer Vision, volume I, pages 109--122, Copenhagen, Denmark, May 2002.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....indicates that a pure bottom up approach is inappropriate. In fact we believe that higher level processes related to recognition should participate in the segmentation process. This idea is reflected in the works of several researches. For a recent non PDE approach starting from such viewpoint see [1]. In this paper, we combine both data driven and recognition driven processing in an unbiased way by introducing a dynamic labeling function into a variational segmentation approach with shape priors. In analogy to the reasoning of Mumford and Shah, minimization of the proposed functional with ....

E. Borenstein and S. Ullman. Class-specific, top-down segmentation. In A. Heyden pages 109--122, Copenhagen, May 2002. Springer, Berlin.

....in the background of the object, scale normalization of the training sample) should be reduced to a minimum or elimi nated. The problem of describing and recognizing categories, as opposed to specific objects (e.g. 6, 9, 11] has recently gained some attention in the machine vision literature [1, 2, 3, 4, 13, 14, 19] with an emphasis on the detection of faces [12, 15, 16] There is broad agreement on the issue of representation: object categories are represented as collection of features, or parts, each part has a distinctive appearance and spatial position. Different authors vary widely on the details: the ....

E. Borenstein. and S. Ullman. Class-specific, top-down segmentation. In Proc. ECCV, pages 109 124, 2002.

....motion etc) boundary smoothness and continuity are used to detect perceptually coherent units. Segmentation can also be performed in a top down manner from object models, where object templates are projected onto an image and matching errors are used to determine the existence of the object [1]. Unfortunately, neither approach alone produces satisfactory results. Without utilizing any knowledge about the scene, image segmentation gets lost in poor data conditions: weak edges, shadows, occlusions and noise. Missed object boundaries can then hardly be recovered in subsequent object ....

....object recognition process. It learns classifiers from training images to detect parts along with the segmentation patterns and their relative spatial configurations. A few approaches based on pattern classification have been developed for part detection [9, 3] Recent work on object segmentation [1] uses image patches and their figure ground labeling as building blocks for segmentation. However, this is not the focus of our paper. 2) Bottom level: pixel based segmentation process. This process finds perceptually coherent groups using pairwise local feature similarity. 3) Interactions: ....

E. Borenstein and S. Ullman. Class-specific, top-down segmentation. In European Conference on Computer Vision, 2002.

....motion etc) boundary smoothness and continuity are used to detect perceptually coherent units. Segmentation can also be performed in a top down manner from object models, where object templates are projected onto an image and matching errors are used to determine the existence of the object [1]. Unfortunately, either approach alone has its drawbacks. Without utilizing any knowledge about the scene, image segmentation gets lost in poor data conditions: weak edges, shadows, occlusions and noise. Missed object boundaries can then hardly be recovered in subsequent object recognition. ....

....object recognition process. It learns classifiers from training images to detect parts along with the segmentation patterns and their relative spatial configurations. A few approaches based on pattern classification have been developed for part detection [9, 3] Recent work on object segmentation [1] uses image patches and their figure ground labeling as building blocks for segmentation. 2)Bottom level: pixel based segmentation process. This process finds perceptually coherent groups using pairwise local feature similarity. The local features we use here are contour cues. 3)Interactions: ....

E. Borenstein and S. Ullman. Class-specific, top-down segmentation. In European Conference on Computer Vision, 2002.

....general images. However, our approach is limited to capturing styles that are local, and at the fragment level, the model is simple and fixed. 2. Previous work Many operations ranging from low level vision tasks to high end graphics applications have been efficiently performed based on examples [3, 6, 8, 9, 10, 15, 16, 20, 21, 27, 29]. The idea of defining and factoring image style and content using a bilinear model was introduced to computer vision by Freeman and Tenenbaum [17] Given a training set of aligned face images of different people under varying illumination, Tenenbaum and Freeman [29] refer to the identity of a ....

....taking its segmentation map as input allows the texture to be painted by numbers [18] A new image that is composed of the various textures is synthesized by painting a new segmentation map. Swapping the output segmentation map with the original input image results in example based segmentation [8]. In our case, we cannot only approximate a common representation for images by blurring and median filtering and then proceed by analogy, as this results in the loss of image detail. Freeman et al. 15, 16] derive a model used for performing super resolution by example. The technique is based on ....

E. Borenstein and S. Ullman. Class-specific, top-down segmentation. In European Conference on Computer Vision, pages 109--124, 2002.

....is used with a deformation space modeled from training data. Some well known applications are: detecting the eye and mouth [11] tracking shapes in motion [1] and segmenting anatomical parts in medical images [5] An alternative to deformable templates for object segmentation is proposed in [2]. Instead of a globally constrained template, object knowledge is represented using pairs of image fragments and their figure ground labeling from a training set. An energy function is formulated for segmenting a test image so that it can be covered by a set of fragments whose appearances match ....

....these top down object segmentation methods require image data to conform to object models, whether encoded in templates or fragments. Here, adopting image patches as a representation, we propose a parallel segmentation and recognition system that also addresses the above mentioned shortcomings of [2]. Our basic idea is that image segmentation should take both low level feature saliency and high level object famil iarity [6] into account. With the guidance of object knowledge, segmentation would not get lost in imaging noise and background clutter. With the verification of low level feature ....

E. Borenstein and S. Ullman. Class-specific, top-down segmentation. In European Conference on Computer Vision, 2002.

....motion etc) boundary smoothness and continuity are used to detect perceptually coherent units. Segmentation can also be performed in a top down manner from object models, where object templates are projected onto an image and matching errors are used to determine the existence of the object [1]. Unfortunately, neither approach alone produces satisfactory results. Without utilizing any knowledge about the scene, image segmentation gets lost in poor data conditions: weak edges, shadows, occlusions and noise. Missed object boundaries can then hardly be recovered in subsequent object ....

....object recognition process. It learns classifiers from training images to detect parts along with the segmentation patterns and their relative spatial configurations. A few approaches based on pattern classification have been developed for part detection [9, 3] Recent work on object segmentation [1] uses image patches and their figure ground labeling as building blocks for segmentation. However, this is not the focus of our paper. 2)Bottom level: pixel based segmentation process. This process finds perceptually coherent groups using pairwise local feature similarity. 3)Interactions: linking ....

E. Borenstein and S. Ullman. Class-specific, top-down segmentation. In European Conference on Computer Vision, 2002.

....segment shred S # i that we wish to find, c e) a few representative image segment shred pairs from the training image. The best matching image shred in this case is (e) to separate the foreground from the background, this representation reduces to the one used by Borenstein and Ullman [1]. However in the more general case of an image containing multiple segments, we will show that this representation can also be considered as a connectivity pattern in a graph. The process of segmentation for a test image subsequently breaks down into two stages: matching and assembly. 2.1 ....

....a contour boundary and partitioned the test image correctly. In both examples, the window size was set to 11 11 (r = 5) The number of components k for clustering was set to 4 and 7, respectively. 4 Related Work Most closely related to our approach is the recent work of Borenstein and Ullman [1] which uses hand labelled image fragments to stitch together figure ground maps for images of horses. Their work, however, does not generalize to more than two segments (foreground background) and they use a greedy algorithm to resolve conflicts between the labels in overlapping fragments. The ....

[Article contains additional citation context not shown here]

E. Borenstein and S. Ullman. Class-specific, top-down segmentation. In Proc. 7th Europ. Conf. Comput. Vision, May 2002.

No context found.

E. Borenstein and S. Ullman. Class-specific, top-down segmentation. In Proc. 7th European Conf. on Computer Vision, volume I, pages 109--122, Copenhagen, Denmark, May 2002.

No context found.

E. Borenstein and S. Ullman. Class-specific, top-down segmentation. In Proc. 7th Europ. Conf. Comput. Vision, volume 2, pages 109--124, 2002.

No context found.

E. Borenstein and S. Ullman. Class-specific, top-down segmentation. In Proc. 7th Europ. Conf. Comput. Vision, volume 2, pages 109--124, 2002.

No context found.

E. Borenstein and S. Ullman. Class-specic, top-down segmentation. In ECCV 2002.

No context found.

E. Borenstein and S. Ullman. Class-specific, top-down segmentation. In ECCV'02, 2002.

No context found.

E. Borenstein and S. Ullman. Class-specific, top-down segmentation. In ECCV'02, LNCS 2353, pages 109--122, 2002.

No context found.

E. Borenstein and S. Ullman. Class-specific, top-down segmentation. In Proc. European Conf. Computer Vision, pages 109--124, 2002. Timor Kadir et al.

No context found.

E. Borenstein and S. Ullman. Class-specific, top-down segmentation. In Proc. European Conf. Computer Vision, pages 109--124, 2002. Timor Kadir et al.

No context found.

E. Borenstein and S. Ullman. Class-specific, top-down segmentation. In ECCV'02, LNCS 2353, pages 109--122, 2002.

No context found.

E. Borenstein and S. Ullman. Class-specific, top-down segmentation. In A. Heyden pages 109--122, Copenhagen, May 2002. Springer, Berlin.

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