| 摘要 |
A method for scaling images by a factor of two. Two variables are extracted from an unaligned input stream. The variables are selectively interleaved. The high and low order bytes of the results of the interleaving are multiplied to convert to a 16-bit representation. Horizontally interpolated data is generated by adding pairs of this conversion. The previous iterations of horizontally interpolated data is retained in buffers. Vertically interpolated data is formed by adding current and previous horizontally interpolated data together. The horizontally interpolated data is then doubled to maintain equal precision between the vertically and horizontally interpolated data. A packing function is employed to convert the data back to an eight-bit representation for output to the destination. Because the precision is the same, a single implicit divisor in the packing instruction handles both horizontally and vertically interpolated data.
|
