Fast Eulerian Filters for Image Interpolation and Representation

摘要

This paper presents image interpolation and approximation techniques under a multiresolution frame work. Different Eulerian filters have been derived from the derivatives of a sigmoid activation function extensively used in artificial neural networks (ANN). The higher order filters can be factorised and expressed in a product form. Each basic filter can be easily implemented recursively with three multiplications and additions only. Frequency response characteristics for the pre-filters and their corresponding interpolators are presented to reveal the quality of interpolation. Some experimental results on image data compression are produced to justify the use of the proposed scheme.