基于加权ENO的图像放大算法

Image Zoom Algorithm Based on Weighted ENO

F.Arandiga的图像自适应插值方法在图像边界区域使用了ENO方法进行插值。通过比较差商的绝对值的大小自适应地选择模板,尽量避免所选择的模板中包含间断,有效地抑制了Gibbs振荡,但仍有很多不足。为弥补ENO方法的缺点,提高插值方法的精确度,提出基于加权ENO的图像放大方法。基本思想是将图像的离散化形式看成图像在单元网格上的平均值,先判断每个单元中是否存在图像边界,在图像边界区域使用加权ENO方法插值图像,在光滑区域使用线性平均插值。该放大方法能得到比F.Arandiga的图像自适应插值方法更高阶的精度。

Adaptive interpolation of images presented by E Arandiga uses Essentially Non-Oscillatory（ENO） schemes to interpolate close to an edge. The key idea of ENO schemes relies in the adaptive stencil which automatically selects the interpolating values in the locally smooth region. It can suppress the Gibbs oscillation effectively but still has several limits. To obtain higher accuracy, image zoom algorithm based on weighted ENO is proposed. The digitalization of an image involves an averaging process and it is appropriate to consider that the pixels of the image represent the cell average of a function. The idea is also an adaptive interpolatory procedure. And in smooth regions the linear average interpolation is used, while close to the edge the weighted ENO schemes are used to interpolate it. This weighted ENO scheme is constructed and has the same stencil nodes as the ENO but can have higher order accuracy. Experimental results demonstrate that the weighted ENO scheme can keep edges better than the ENO scheme.