摘要 |
PROBLEM TO BE SOLVED: To process the pixels of all output images without operating any complicate condition branch, and to execute a plurality of instructions in parallelism by the parallelism of an algorithm. SOLUTION: A parameter (size or position) related with the patch of an output image is calculated, and the coordinate mapping of the pixels of an output image and the pixels of an input image is executed so that a patch table can be prepared. Then, the region of the output image is divided into rectangular patches based on the patch table. Furthermore, the patch is divided into pixel lines, and the pixel line is divided into pixels. The pixels are decomposable into four components (C, M, Y, K), and the four components of the corresponding input image are simultaneously read (S604). In this case, conversion processing is executed to each component (S605), and totally the same conversion processing is executed to those four components so that parallel processing can be executed by using an SIMD instruction. Also, a plurality of instructions are executed in parallel without operating any complicate condition branch. At last, the components of the pixels of the output image are simultaneously written (S606). COPYRIGHT: (C)2003,JPO
|