| 摘要 |
PURPOSE:To obtain an inexpensive and high-speed operator using one multiplier, by stopping the multiplication of upper bits to perform the multiplication of only lower bits when upper bits of correction data become too small to have no influence upon the result of the product sum. CONSTITUTION:In a convolution operator which operates the product sum between measurement data and correction data for image reconstitution in a CT device using transmission X rays, ultrasonic waves, or the like, the number of bits of correction data for image reconstitution is made larger than that of measurement data, and the number of bits of this correction data is divided to upper bits and lower bits to perform the multiplication. When upper bits of correction data for image reconstitution become too small to have no influence upon the result of the product sum of said multiplication, the multiplication of upper bits is stopped, and only the multiplication of lower bits is performed. For example, the output of the second memory 22 for weight data is inputted to one multiplier 23A through an upper/lower selecting circuit 26 to perform said operation. |
