The adjacent diffusion filter for image enlargement

Image enlargement has widely been used in many fields, such as image display, video effects etc. When a good enlargement quality and a simple algorithm are expected simultaneously, the traditional methods without considering the directional sensitivity of human vision runs into difficulty. In this paper, the adjacent diffusion - a new method that matches human vision for image zooming is proposed. By means of a simple coordinate transformation, which especially benefits for hardware implementation, a fairly good quality can be obtained. In this paper, a new 2-dimension lowpass filter - adjacent diffusion filter and its frequency response are analyzed, and the diffusion increment matrices for the actual application of the filter are constituted, then the coordinate transformation for an arbitrary (N/M)×(N/M) enlargement is given. The results of this paper including some examples can also be regarded as experimental evidence on the directional characteristic of human vision.