The source block is shape compensated (filtered) to look like the target block. The frames are further divided into blocks of 16 �� 16 pixel size. We then define the pattern with size of 3 �� 3 with its center as the working pixel on the source block. Thus, there are 16 �� 16 moving patterns in each scanned block. Patterns of the same kind are grouped together as pattern group. Therefore there are at most 512 pattern groups. We need to select a filter for each source block. The filters for selection are candidates. Every pattern in the block is a voter and a pattern group is a voter group. Thus, there are 16 �� 16 voters of at most 512 voter types for every block. We can prepare the filter-pattern relation table off-line by filtering every type of pattern. One typical table composed of 256 filters is shown in Table 1.

Table 1.Filter-pattern relation table.We then define the target pixel on the target block by the corresponding pixel with the same location of the center of the pattern (voter) on the source block. For each pattern (voter), the filtered result is either consistent with the target pixel or not. The optimal filter is the filter causing the least inconsistent results. In other words, the candidate on selection is the candidate accepted by the most voters. The optimal filter (winner) can be selected pattern (voter) by pattern (voter) or group by group. Group is short for pattern (voter) group. Selection by group is most advantageous in the situation where the voters are much more than the groups and is occurred during the off-line selection in the next section.

The pattern group associated with the target value is called pattern-target relation or pattern-target occurrence table. One realization of this relation is shown in Table 2.Table 2.Pattern-target occurrence table.In practice, we first scan the Anacetrapib source and target blocks to have Table 2 (pattern-target occurrence table). Then, we build a pattern-filter conflict table from Table 1 (the off-line prepared pattern-filter relation table) and Table 2 (pattern-target occurrence table). Using this table, the least inconsistent filter is obtained. One example of the pattern-filter conflict table deduced from Table 1 and Table 2 is shown in Table 3.Table 3.Pattern-filter conflict table.We summarize the processing procedures as follows:Step 0.

Off-line preparing the filter collection (explained in next section) and Table 1 (Filter-pattern relation) table.Step 1. Building Table 2 (Pattern-target occurrence table) in a single scan of the corresponding source and target blocks.Step 2. Building Table 3 (Filter-pattern conflict table) from Table 1 and Table 2 by checking the target value in Table 2 with the relation value in Table 1.Step 3. Finding the least conflicts filter from Table 3 by summing the column.4.