Integer smoothing filters in residue number system

In this paper we propose a new method for image smoothing using the Residue Number System (RNS). The essence of the approach is to replace the computationally complex division operation in the RNS by multiplication of all the fractional numbers by a power of two and then rounding them off. As a result, all subsequent calculations are performed only on fixed-point numbers. The theoretical and practical studies were performed, which have shown that the rounding error does not affect the result of image filtration, if a certain accuracy of calculations is reached. In addition, there are possibilities for the balance control between the accuracy of calculations and the dimensionality of variables, which opens up the potential for effective hardware implementation on the FPGA.