"... In PAGE 3: ... Generally, the ultimate tn k are taken to be very large to permit all possible region merges. Table1 shows the organization of the parameters of the various segmentation levels. Note that each level is created by the application of multiple merge predicates to pairs of adjacent regions.... In PAGE 4: ... than regions with a common quadtree parent as in the initialization). Adjacent regions Rl i; Rl j are merged into one region whenever the predicate P l k is true, where, given a cost-of-merging function Cl k, P l k(Rl i; Rl j) Cl k(Rl i; Rl j) lt; tl k : Multiple merging predicates may be successively applied at each level, as in Table1 . Our criteria are derived from simple image statistics, e.... ..."

"... In PAGE 9: ... These are presented so that the performance of our technique can be evaluated in the context of other dithering techniques. Table1 describes the parameters used for the display of each of the sample images. Examining the two-colour approximations we notice some step-like e ects.... In PAGE 24: ...Table1 for the parameters used in this gure)... ..."

"... In PAGE 8: ... Reporting at the regional scale The next issue concerns the appropriate scale for reporting landscape pattern indicators for a hetero- geneous area such as Region IV. Table2 gives means and standard deviations for the three subre- gions. These values can be compared to the indica- tors calculated for the entire region.... In PAGE 9: ... Notice that, on average, Subregion C has many more different kinds of edges or adjacencies, indi- cating a landscape with many landcover types interwoven into a complex pattern. The final observation about Table2 is that the grand mean for the entire region tends to lose the distinctions among subregions, providing an inter- mediate average value. One would expect that large, undesirable landscape changes could occur in subregions A or C without detecting any signifi- cant change in the grand mean.... ..."

"... In PAGE 8: ... Figure 3a shows the empirical histogram of the 5,303 distinct tags represented by 19,527 tags in the yeast library derived from G2/M phase-arrested cells, and of the 3,200 true distinct tags of the same library after the elimina- tion of 2,103 distinct tags associated with 3,239 tags that match noncoding genomic regions and antisense sequences. Most of these erroneous tags occur with only 1 or 2 copies ( Table2 , Figure 3b). These erroneous tags comprise 16.... In PAGE 8: ...2% of library size (Table 2). Figure 3 and Table2 show the GDP model at b = 0 (simple power law) fitted well a frequency distribution of different classes of erroneous tags, but b gt;0 in the case of frequency distribution of true tags. Matches of many distinct tags to the same gene, and one distinct tag to many genes, constitute serious and com- mon problems in correctly identifying genes and properly determining their expression levels [5, 26, 29], particularly in larger SAGE libraries.... ..."

"... In PAGE 27: ... The 36 cases were obtained by forming all combinations of the 4 geometries, the 3 constraint sets, and the 3 maximum beam set sizes. Table3 summarizes the performance results obtained. Each of the 4 major columns refer to one of the 4 geometrical cases.... ..."