| （英） |
In prior research, we verified carrying out experiments on image processing programs, considering algorithms and architectures for large-scale parallel computing. Indeed, while a certain degree of parallel computing effectiveness was observed in patterns with many loop operations, the speedup of the image processing programs varied. Therefore, when assigning computationally intensive tasks to a computer, it is considered possible to reduce computational load and achieve efficient processing by pre-computing data that can be calculated in advance, storing these values in an array (LookUp Table), and retrieving the desired data from the array instead of performing calculations each time. Theoretically, replacing expensive computational processing or I/O processing with a lookup table can significantly reduce processing time. However, since verification using image processing programs was insufficient, it was deemed necessary to clarify this. In this study, first multiple image processing programs were prepared. Subsequently, verification experiments were conducted on the image processing programs, considering lookup table algorithms and architecture, with the goal of accelerating the programs. The effects of the algorithms and architecture were then examined. |