| 摘要 |
A novel method is disclosed for laying out a plurality of rectangles onto a bounding box area, where bboxArea represents a total area of the bounding box and totalRectArea represents the sum of the areas of all the rectangles. The method comprises the steps of: first adjusting all rectangles to a specified aspect ratio and then computing a sum of areas of all rectangle intersections (overlap) occurring as a result of said arrangement. Next, the function blackArea=totalRectArea-overlap is used to approximate the black area and all non-overlapping spaces remaining between the arranged rectangles are computed by: WhiteSpace=bboxArea-blackArea. Then, for the current arrangement, an energy function E=whiteSpace+(overlapPenlalty * overlap) is calculated and for each energy function, a state probability function is calculated such that: Pr=exp(-(E<SUB>new</SUB>-E<SUB>old</SUB>)/kT), where E<SUB>old </SUB>is the energy computed for the previously accepted state, E<SUB>new </SUB>is the energy calculated for the current state, T is a control parameter from T<SUB>0</SUB>-T<SUB>end</SUB>, and k is a constant. Typically, initial value of T<SUB>0</SUB>=100 and T<SUB>end</SUB>=0.0 and the overlapPenalty has an initial value of 100. The value for the control parameter T is subsequently lowered by a fraction of its present value using the relationship: T<SUB>N+1</SUB>=T<SUB>N</SUB>* 0.95. A random number between 0 . . 1 is then selected and if the random number is less than the value of Pr then the new state E<SUB>new </SUB>is accepted thus new states are always accepted where E<SUB>new </SUB>is less than E<SUB>old</SUB>. The rectangles are again randomly arranged on the bounding box area and the method repeats itself until either T<SUB>end </SUB>is reached or the value of Pr is within acceptable limits.
|
