"... In PAGE 13: ... Though in general L is larger than n or even n 2 , the running time of MRAM method in average is less than other three methods because of its multi-resolution strategy. Table3 shows the relative similarity query processing time of the four methods using the simple system we described in [16]. The experiments are conducted on a SunSparc workstation... ..."

"... In PAGE 16: ....mn/ respectively. Though in general L is larger than n or even n2, the running time of MRAM method in average is less than other three methods because of its multi- resolution strategy. Table3 shows the relative similarity query processing time of the four methods using the simple system we described in [17]. The experiments are conducted on a SunSparc workstation 10/30.... ..."

"... In PAGE 22: ... The selection of national level indicators may focus on processes that shape these variations in time and space. Some examples of processes that are relevant to adaptive capability are listed in Table1 . It has been argued, for example, that traditional systems of social security in Third World societies may be in the process of being eroded due to the joint impact of market penetration (reorientation of most production away from local circulation and reciprocity), population growth (fewer unappropriated flexible resources) and rise of the modern state system (provides services) (Platteau 1991).... ..."

"... In PAGE 5: ... 3.3 Time required for Searching 3D Models Table1 shows the time required for the processing of 1787 3D model data by using a Pentium IV 3.4Ghz with 1024 MB of main memory.... In PAGE 5: ... This process occasionally slows down decomposition time. All the processes in Table1 are pre-processes of the similarity retrieval and have to be done only once. The system might require a considerable amount of time for initial program code execution for a similarity search because the system has to open/close files for read- ing shape descriptors 52274 times.... ..."

"... In PAGE 25: ... In this section features of existing CSCW applications are identified that have to be tested empirically, in order to ensure user-centered development. Table6 lists the features along the CSCW categories as introduced in section 2 and table 2 and 3. Subsequently, the findings of Table 6 are commented for each type of awareness.... In PAGE 25: ... Table 6 lists the features along the CSCW categories as introduced in section 2 and table 2 and 3. Subsequently, the findings of Table6 are commented for each type of awareness. Group-structural awareness is not considered due to the lack of features.... ..."

"... In PAGE 7: ... The process was repeated 20 times. Comparative classification performance for individual class can be seen in Table2 . In this experiment, the clas- sification accuracy was lower than the first experiment due the reasons, (1) Shape samples within the class show larger variation, as shape of vehicles of different makes and mod- els vary.... ..."

"... In PAGE 18: ... The large number of small faces therefore leads to a rather large 203% overhead for the patch_borders component. Table2 shows the processing times of the three sub-phases (the initial phase is mostly due to the initialization of the bucketing data structure, sam- pling and texture packing) when various attributes are preserved: Color; shape detail in the two modal- ities, normal maps (Shape1) or displacement maps (Shape2); or both the previous detail integrated into a preshaded color texture (Both1, Both2). The mesh used in Table 2 is fandisk, simplified to 100 faces from the original 12 946.... In PAGE 18: ... Table 2 shows the processing times of the three sub-phases (the initial phase is mostly due to the initialization of the bucketing data structure, sam- pling and texture packing) when various attributes are preserved: Color; shape detail in the two modal- ities, normal maps (Shape1) or displacement maps (Shape2); or both the previous detail integrated into a preshaded color texture (Both1, Both2). The mesh used in Table2 is fandisk, simplified to 100 faces from the original 12 946. Table 2 also highlights the fact that the only phase that depends on texture res- olution is the sampling time.... In PAGE 18: ... The mesh used in Table 2 is fandisk, simplified to 100 faces from the original 12 946. Table2 also highlights the fact that the only phase that depends on texture res- olution is the sampling time. Producing shape detail via displacements is slightly more expensive because we first need to evaluate displacements and then we compute each normal from the displacements around each sampling point.... ..."

"... In PAGE 2: ... The half re-entrant shape has an improved technical advantage of easing the fluid flow and avoiding gas pockets during the cavity processing, as com- pared to the re-entrant shape. RF parameters of the TTF shape and the new shapes are summarized in Table1 . In summary, as compared to the TTF shape, the new shapes lower the H pk =E acc by 9-15% at the price of a 10-35% higher E pk =E acc .... ..."