基于小波变换的图像超分辨率复原算法研究

Study of Super-resolution Image Restoration Algorithm Based on Wavelet Transform

在不改变现有硬件设备的情况下,结合近年来迅速发展的小波理论,提出了基于小波变换的图像超分辨率算法。对输入的低分辨率图像采用直接邻域进行插值后,利用DWT将低分辨率图像分解为不同的4个子带;同时直接对低分辨率图像进行SWT处理。由SWT得到的高频频带来修正DWT得到的高频频带,可修正估计系数。最后,通过逆离散小波变换(IDWT)组合修正的高频频带和输入图像,得到一幅高分辨率的输出图像。实验证明,与传统的双线性插值、双立方插值相比,该算法的峰值信噪比PSNR都有不同程度的提高。

Under the circumstances of without changing the existing hardware device,and taking into account that wavelet theory had been developing rapidly in recent years,this paper presented the image super-resolution algorithm based on wavelet transform.After direct neighborhood interpolation,low resolution image will be decomposed into four different Sub-band with DWT and at the same time is directly processed by SWT.High frequency band will be amended from using SWT to DWT,so as to fix the estimated coefficient.Finally,a high resolution output image can be obtained with high frequency band and the input image being modified by inverse discrete wavelet transform（IDWT）.Experiments show that compared with the traditional bilinear interpolation and bicubic interpolation,the peak signal-to-noise ratio PSNR of the algorithm presented in this paper is improved.