Image interpolation using mathematical morphology

TLDR: A new method for interpolating binary images that outperforms existing techniques, based on mathematical morphology, a theoretical framework to alter an image while preserving the image objects' geometry.

Abstract: We present a new method for interpolating binary images that outperforms existing techniques. Bitmapped images have a specific horizontal and vertical resolution. When we magnify such an image, we want the resolution to be increased, allowing more details in the image. However, these extra details are not present in the original image. A blowup of the image using simple interpolation will introduce jagged edges, also called "jaggies". We present a new interpolation technique "mmINT", which avoids these errors. It is based on mathematical morphology, a theoretical framework to alter an image while preserving the image objects' geometry. The algorithm detects jaggies in the blown up image and removes them, making the edges smoother. This is done by replacing specific black pixels with white pixels, and vice versa. The results show that mmINT is a superior technique for the interpolation of binary images, like logos, diagrams, cartoons and maps.