Research On Image Scaling Algorithms

Image scaling plays a very important role in image processing. Image scaling technique is used in image resolution adjustment and format conversion. With the development of multimedia technology and large-scale integrated circuits, implementation of free scaling in hardware becomes an important research topic. Generally we use interpolation to scale an image.Traditional polynomial-based intepolation methods apply the same kernel on the whole image, with the coefficients predetermined and not related to the content of the image, thus the details and edge of the image are smoothed in the intepolated image and the quality of the image is degraded. To solve this problem, there are many edge adaptive interpolation methods in literature. However, most of these methods are too complex and not efficient enough to implement in hardware.To solve the problems discribed above, this paper has done several jobs as follows:1. Image interpolation theory and all kinds of polynomial-based interpolation and approximation methods are discussed systematically;2. Edge detection operators and several classic edge adaptive interpolation methods are addressed. We find that ALC algorithm(adaptive low complexity algorithm for image zooming) is very good for texture region. The performance of ALC in texture region is much better than traditional algorithms according to the experimental results.3. Improved edge adaptive interpolation(IEAI) algorithm has been proposed. This algorithm has two improvements:determination of edge direction has been simplified, this decrease computation dramatically; cubic B-spline approximation has been used for flat region, this method can inhibit gauss white noise.4. Performance of several interpolation methods has been compared and analized based on Fourier analysis and subject evaluation.