A novel algorithm for digital image processing applications

This paper introduces an efficient algorithm for the calculation of 2-D convolution and correlation for image processing applications. The method combines the new 2-D Mersenne with the 2-D Fermat numbers transforms using the 2-D mixed radix conversion. The moduli of the transforms are selected to be close to one another, thus, the choice of the dynamic range and the constraint between the transform sizes and the world length becomes more flexible. The residue transforms are independent and can be calculated in parallel for high speed and high throughput. Hence the technique is suitable for image processing applications such as error free image filtering and enhancement.