| 摘要 |
PURPOSE:To shorten the processing time and to prevent the contents of an operation from being limited by bringing selectively an arbitrary point to convolution processing in accordance with the contents of the operation in a matrix of NXN pieces, which becomes an object of the operation. CONSTITUTION:An arithmetic order signal 2 generated from a sequence counter 1 is applied to an arithmetic point RAM 3 and a coefficient memory 9, and from the arithmetic point RAM 3, the corresponding arithmetic object point address signal 4 is outputted, becomes an address input of an image memory 6, arithmetic object point data 8 is outputted, and from the coefficient memory 9, the corresponding convolution coefficient signal 11 is outputted, multiplied by a multiplier 12, and accumulated by an accumulator 14. The RAM 3 outputs a sequence control signal 5, and in the case the arithmetic object point is ended, it is informed to a counter 1, and also, a result of accumulation is outputted as a processing result signal 15. |
