"... In PAGE 1: ... On the world-wide web (WWW), various experi- mental approaches are already available[9], from which some lessons can be drawn. Textual methods which are based on keyword queries ( Table1 , A) to nd images are potentially very powerful. However, a textual annotation of images produced by a single content provider, al- though already very costly by itself, usually does not cover a su cient number of views or perspec- tives on the same pictorial material.... ..."

"... In PAGE 18: ... Such an implementation of cut-and-paste operates universally, irrespective of the particular implementation or representation of content that an application is using2. A comparison of capabilities of structured and image-based document representations is shown in Table1 . We note a number of advantages of image-based representations over structured representations of documents: more intuitive behavior, ability to represent arbitrary content, simple format conversions, lower up-front investment in markup, and improved privacy and security.... ..."

"... In PAGE 9: ... Similarity measures of all the articles in this corpus with the respective four reference articles were made using the image-based and text-based methods. The results are summarized in Table1 and Figure 4. The above images came from paper printed by a printer.... ..."

"... In PAGE 5: ... It changes the coordinates of the centroids in the table by the correspondent value of the zone of influence image at that coordinate. Table4 and Table 5 show the correspondent connected tables of Table 1 and Table 2 converted to the Reference Holes Regions. This operator can check for two possible errors: a) the existence of two or more holes in same zone of influence; b) the lack of hole in any zone of influence.... In PAGE 5: ...Table 5 - Connected Table of the Test image based on the reference zone of influence. This operator is explained using the illustrative example of the Table4 and Table 5, which refer to the images of Figure 3: 1. Obtain the number of conductors of the Test table.... ..."

"... In PAGE 24: ... Of course, since any two points on the object model can serve as compile-time parameters to a scaling TP, many other parameterizations of the scaling TP could be included in the library. Although the visual procedure library shown in Table1 is su cient for the pur- poses of this experiment, it includes just a few of the computer vision algorithms described in the literature. Unfortunately, the current Lisp implementation of SLS has proved an impediment to building a larger library, since source code for most visual procedures is available only in C.... ..."

"... In PAGE 6: ... [13]; and (2) a recent version of the NCuts segmentation algorithm [6]. In Table3 , we show that our algorithm achieves greater mean conservation and efficiency than both of the others. Also, only our algorithm provides figure/ground labels.... ..."

"... In PAGE 9: ... In contrast with 3-D model-based rendering, image-based rendering tech- niques rely primarily on the original or trained set of images to produce new, virtual views. Comparisons between the 3-D model-based rendering and the image-based rendering tech- niques are shown in Table1 . In 3-D model-based rendering, 3-D objects and scenes are repre- sented by explicitly constructed 3-D models (from CAD modeler, 3-D digitizer, active range, or stereo techniques).... ..."

"... In PAGE 3: ... Character accuracy is defined as: Character Accuracy = (n - Number of Errors) / n where n is the total number of characters in the ground- truth text [1]. 5: Results and Analysis Table1 and 2 summarize the decisions made by the image- based prediction system processing the Sample 2 data set and the magazine data set. For all systems, this classifier was able to correctly recognize approximately 85% of the pages in each data set.... ..."

"... In PAGE 3: ... This imposes severe restrictions on the freedom of choosing novel viewpoints. Table1 summarizes the most relevant work and compares them with our proposed new method. For more comprehen- sive surveys on the reconstruction of visual hulls and photo hulls, please be referred to [Dye01, SCMS01].... ..."

"... In PAGE 11: ...s many as possible news articles, then a threshold of 0.1 may be adopted. Setting 0.15 as the threshold appears a good compromise. Table1 also shows a slightly better performance for English imaged documents than Chinese, the reason being the more complicated character shapes in Chinese and hence a much greater number of character object classes to handle. 0 0.... ..."