多小波图像插值算法研究

与单小波变换一样,多小波变换同样具有多分辨分析的特性,1次多小波变换可以将图像分解成4个低频子带和12个高频子带,而且原图像的大小是每个子带的4倍.根据多小波变换的这一特点,利用原图像与经过1次多小波变换后的各高频子带的信息,并考虑各子带的分形维数,提出了一种新颖的灰度图像插值算法.实验结果表明,与传统的插值算法相比,例如双线性插值与双三次多项式插值,该算法的插值效果较好,且克服了单小波插值中出现的斑点干扰.

Study to the Image Interpolation Algorithm Based on Multiwavelet Transform

Similar to the scalar wavelet transform,the multiwavelet transform also have multiresolutuion characterristics.By one level multiwavelet decomposition,the image is transformed into four low frequency bands and twelve high frequency bands,and the size of original image is four times larger than that ones in the transformed bands.With this characteristics of multiwavelet transformations,and making use of the information obtained from the original image and the high frequency bands after transformation,also taking the fractal dimensions of high frequency bands into account,the paper has proposed a novel image interpolation algorithm for grey images.The experimental results shows that the new algorithm has overcome shortcomings of the blur effects when interpolating with the scalar wavelet transform and has better results comparing with the traditional image interpolation methods such as bilinear or bicubic ones.