基于整数小波变换的Wiener插值算法

Wiener interpolation methods based on integer wavelet transform

小波变换中高低分辨率子带之间的相似性使得利用小波变换进行图像插值的方法成为可能。根据数字图像信号的特点,分析了小波变化后各个子带信号的特点,提出了基于整数小波变换的Wiener插值算法。用Wiener自适应滤波器训练得到插值滤波系数,同时结合既符合图像性质又能减小运算量的整数双正交小波基对图像插值。结果得到较高的信噪比和较好的主观视觉效果。平均峰值信噪比比传统的双线性插值法提升了2.4dB。

Based on different models, several image interpolating algorithms such as bilinear interpolation, spline interpolation, etc. have been developed. Moreover, due to the property of the similarity between high and low resolution image sub-bands in wavelet domain, it is possible to bring forward a new image interpolating method based on wavelet transform. In this paper, by considering the characteristic of digital image signals and analyzing each sub-band signal in wavelet domain, a Wiener interpolation method based on integer wavelet transform was introduced. We use adaptive filters to train a group of coefficients for interpolating and an integer bi-orthonormal wavelet base to do the inverse transform with interpolated sub-band signals, which can be compatible with image attributes and also reduce the computational complexity. As the experiment results, our method has a higher PSNR and a better subjective quality. Comparing to common used interpolation methods, the average PSNR gain of our method is 2.4dB.