| 摘要 |
PURPOSE:To reduce multiplication frequency and input frequency and to attain rapid filtering processing by executing the processing of respective lines in parallel in the filtering processing and dividing the line processing itself into parallel parts. CONSTITUTION:In order to read out data in each line from an image memory and execute line processing in parallel, an input processing part 1 distributes data to three processors 101 to 103. In the distributing method, lines 111 to 113 are distributed to the processors 101 to 103 and the sum of products between load data pointed out by the pointer of a load matrix and the data of the line 111 is executed at first. Then, the line 111 of the processor 101 is substituted by the line 114 and the sum of products between the load data pointed out by the pointer indicted by the load matrix and the line 114 is calculated in each processor. Thus, the product sum calculation can be successively executed. A line processing part 2 is driven independently in each processor. Consequently, rapid internal processing can be attained. |
