Z.N. Li , O.R. Zaiane, Z. Tauber, "Illumination invariance and object model in content-based image and video retrieval", J. Vis. Commun. Image Rep., Vol. 10, No. 3, pp. 219-244, 1999.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....pulmonary CT images, and multispectral Landsat satellite images. 6 Figure 5: C bird. 7 C bird (Content Based Image Retrieval from Digital libraries) Developer School of Computing Science, Simon Fraser University, Burnaby, B.C. Canada. URL http: jupiter.cs. sfu.ca cbird References [LZY98] LZT99] Features For each collected image, a feature descriptor and a layout descriptor are computed. A feature descriptor is a set of four vectors: a color vector, a most frequent color (MFC) vector, a most frequent orientation (MFO) vector, and a chromaticity vector. A 512 bin RGB histogram is ....

Z.N. Li, O. R. Zaane, and Z. Tauber. Illumination invariance and object model in content-based image and video retrieval. Journal of Visual Communication and Image Representation, 10(3):219-244, September 1999. http://www.cs.sfu.ca/cs/undergrad/CM/CMPT365/CBS/CBIRD.pdf.

....the ability to automatically locate and track objects in videos is very important. The most intuitive way of accomplishing this task is to first generate temporally tracked homogeneous regions and then apply further processing (automatic or human aided) to identify the semantic objects. In [8], we proposed Feature Localization as an alternative to traditional image segmentation in respect to object based image retrieval. Locales are enclosures of local features that are not required to be connected, disjoint, or complete. Locales also operate at a higher level than pixels, as the basic ....

....philosophy that all calculations are done based on tiles while keeping pixel unit precision; hence we gain the advantage of utilizing both global and local information. Experiments have shown impressive results. 2. FEATURE LOCALIZATION CONCEPT Feature localization was introduced and described in [8]. Definition 1: A locale L f is a local enclosure of feature f. A locale L f uses blocks of pixels called tiles as its building units, and has the following descriptors: 1. Envelope f : a set of tiles representing the locality of L f . 2. Geometry: mass M(L f ) centroid C(L f ) and pixel ....

[Article contains additional citation context not shown here]

Z.N. Li , O.R. Zaiane, Z. Tauber, "Illumination invariance and object model in content-based image and video retrieval", J. Vis. Commun. Image Rep., Vol. 10, No. 3, pp. 219-244, 1999.

....to apply semantic models without higherlevel features than pixels; hence, it is a chicken and egg problem. Typically, precise and accurate segmentation cannot be obtained for general videos. Without region information, it becomes non trivial to extract multiple motions. Li, Drew and Tauber [31, 51, 52] proposed feature localization as an alternative to traditional image segmentation with respect to object based image retrieval. Locales are enclosures of local features that are not required to be connected, disjoint, or complete. Locales also operate at a higher level than pixels, as the basic ....

....of localization is that it generates coarse locales that are more robust to noise and complex object surfaces. It is also much more attainable as it does not require complete pixel level segmentation. Feature localization facilitates the exploration of multiple features. For example, in [31], it is shown that after the initial localization based on color, locale based methods can be developed to greatly improve the quality of edge based texture and shape analysis. Many texture analysis methods are based on statistical measures of edges (e.g. edge density, orientation, separation, ....

[Article contains additional citation context not shown here]

Z.N. Li, O.R. Zaiane, and Z. Tauber. Illumination invariance and object model in content-based image and video retrieval. Journal of Visual Communication and Image Representation, 10(3):219--244, 1999.

....the shape and the motion trails. The algorithm uses temporal consistency to refine the object segmentation. For test queries the precision ranges from 20 100 for recall less than 50 , and is around 25 for higher recall. In our C BIRD (Content Based Image Retrieval from Digital libraries) system [16], we employ the Query by Example regime. Image features are coarsely localized into locales for the purpose of objectbased retrieval. Localization is not segmentation: locales can be overlapped and or non connected, and the set of all locales does not have to include all image pixels. Locales are ....

Z.N. Li, O.R. Zaane, and Z. Tauber. Illumination invariance and object model in content-based image and video retrieval. J. Vis. Comm. and Image Rep., 10:219--244, 1999.

....standard of MPEG 1 and MPEG 2 by allowing the transmission of an arbitrarily shaped video object in one bit stream. The new functionality provides the infrastructure for interactive manipulation of video objects instead of pixels and paves the way for many content based applications, such as [93]. In MPEG 4, there are two types of video objects, opaque and transparent objects. The shape of these objects is represented by a 2D binary mask, where white pixels denote the object and black pixels denote the background. Coding efficiency, scalability, error resilience and hardware complexity ....

Z.N. Li, O.R. Zaiane, and Z. Tauber. Illumination invariance and object model in content-based image and video retrieval. Journal of Visual Communication and Image Representation 10, 1999, p. 219-244.

....objects can greatly affect the global feature vectors. While image segmentation for object representation has been widely studied [4] the problem is arguably ill defined and a good segmentation is often impossible to achieve. In CBIRD (Content Based Image Retrieval from Digital libraries) [5] we developed a technique for coarse localization of image features into locales. The set of image locales may be overlapped and or non connected, and the union of all locales does not have to include all image pixels, i.e. the set may be non disjoint and noncomplete. Since a model object and a ....

....given the estimated pose. In x2, we describe image feature localization, and x3 describes our search method. Section 4 shows results, and x5 states some conclusions. 2. FEATURE LOCALIZATION We have successfully used a coarse localization of image features based on proximity and compactness [5, 7]. A locale Lf uses squares of pixels (tiles) as its basic unit, and has descriptors: 1) envelope Lf the set of tiles for locality Lf ; 2) geometric parameters mass M#Lf #, centroid C#Lf #, eccentricity E#Lf #; and (3) locale texture histogram. Although the envelope Lf of a locale Lf is ....

Z.N. Li, O.R. Zaane, and Z. Tauber. Illumination invariance and object model in content-based image and video retrieval. J. Vis. Comm. and Image Rep., 10:219--244, 1999.

No context found.

Z. Li, O. Zaiane, and Z. Tauber, "Illumination invariance and object model in content-based image and video retrieval," Journal of Visual Communication and Image Representation 10(3), pp. 219--244, 1999.

No context found.

Z.N. Li, O.R. Zaiane, and Z. Tauber. Illumination invariance and object model in content-based image and video retrieval. J. Vis. Commun. & Image Rep., 10:219--244, 1999.

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