Rui, Yong; She, Alfred C. and Huang, Thomas S. (1996). Automated Region Segmentation Using Attraction-Based Grouping in Spatial-Color- Texture Space, to appear in Proc.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....and robust (with respect to image appearance and the required invariance properties) Past research on segmentation for image retrieval includes, on the one hand, feature based methods that focus on the feature space. Among them, we can mention histogram backprojection [18, 16] clustering [14, 20] and neural nets [2] Focusing on the feature space often overlooks the spatial relationships of the pixels in the image space, e.g. there is no evidence that a coherent cluster in feature space corresponds to a coherent region in image space. Thus, feature based methods have to trust the ....

Y. Rui, A. C. She, and T. S. Huang. Automated region segmentation using attraction based grouping in spatial-colortexture space. In Proceedings of the ICIP. IEEE, Oct. 1996.

....optimization problem, which is not feasible in a real time content based retrieval system. Furthermore, to the extent of our knowledge, no research has been done on how to deal with the spatial discretization problem when using these methods. We proposed the Modified Fourier Descriptor (MFD) (Rui, She, and Huang, 1996b) which satisfies the following four conditions: 1. Robustness to Transformation the representation must be invariant to translation, rotation, and scaling of shapes, as well as the starting point used in defining the boundary sequence. 2. Robustness to Noise shape boundaries often ....

....feature is above the threshold, then this region is considered as the object; otherwise, considered as the background. One obvious disadvantage of this thresholding method is that the threshold is image dependent. We propose an attraction based grouping method (ABGM) to overcome this disadvantage (Rui, She, and Huang 1996a) The method is motivated by the way the human visual system might do the grouping. As defined in physics, F 12 = G M 1 M 2 d 2 reflects how large the attraction is between the two masses M 1 and M 2 when they are of distance d. In ABGM, we use the similar concept, but now M 1 and M 2 are ....

Rui, Yong; She, Alfred C. and Huang, Thomas S. (1996). Automated Region Segmentation Using Attraction-Based Grouping in Spatial-Color- Texture Space, to appear in Proc.

....optimization problem, which is not feasible in a real time content based retrieval system. Furthermore, to the extent of our knowledge, no research has been done on how to deal with the spatial discretization problem when using these methods. We proposed the Modi#ed Fourier Descriptor #MFD# #Rui, She, and Huang, 1996b#, which satis#es the following four conditions: 1. Robustness to Transformation # the representation must be invariant to translation, rotation, and scaling of shapes, as well as the starting point used in de#ning the boundary sequence. 2. Robustness to Noise # shape boundaries often contain ....

....feature is above the threshold, then this region is considered as the object; otherwise, considered as the background. One obvious disadvantage of this thresholding method is that the threshold is image dependent. We propose an attraction based grouping method #ABGM# to overcome this disadvantage #Rui, She, and Huang 1996a#. The method is motivated by the way the human visual system might do the grouping. As de#ned in physics, F 12 = G M 1 M 2 d 2 re#ects how large the attraction is between the two masses M 1 and M 2 when they are of distance d. In ABGM, we use the similar concept, but now M 1 and M 2 are ....

Rui, Yong; She, Alfred C. and Huang, Thomas S. #1996#. Automated Region Segmentation Using Attraction-Based Grouping in Spatial-Color- Texture Space, to appear in Proc.

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