Error diffusion is a well known technique for generating bi-level images and is often used instead of a halftone screen process in order to minimize the visual impact of quantization error. In addition, some kernels used for error diffusion are capable of enhancing the appearance of edge sharpness. The current study examined experimentally the effect of kernel type on the magnitude and the directional symmetry of edge enhancement. In addition, an alternative to raster image processing, involving a linear algorithm called linear pixel shuffling, was used to perform error diffusion. This allows the use of symmetrical kernels and the omni-directional diffusion of quantization error. Edge enhancement with omni-directional error diffusion was examined and found to be capable of more directionally uniform enhancement of edges than is possible with raster error diffusion.