一种新的超分辨率图像重建算法

A New Algorithm of Image Super-Resolution Reconstruction

超分辨率重建是提高图像分辨率的一种方法.根据图像分层变换和插值处理的特点,提出了一种将分层子带变换S+P变换(序列变换加预测运算)和插值算法相结合的超分辨率重建算法.该方法首先对图像序列进行运动估计和运动补偿,然后再对配准后的图像序列进行S+P变换,并对分解得到的高频子带进行插值,最后通过S+P逆变换得到超分辨率图像.实验表明,该方法能够较好地保持原图像中的细节,且在主观上具有很好的视觉效果,客观上具有较高的峰值信噪比,是一种有效的超分辨率处理方法.

Super-resolution reconstruction is a kind of method of improving the image resolution. According to characteristics of image stratification transformation and interpolation processing, a new super-resolution reconstruction method based on stratification sub-band S ＋ P transformation （sequence transformation and pre- diction operation） and interpolation algorithm was presented. First, motion estimation and compensation were implemented for image sequence. Then, the registered image sequence was transformed by S ＋ P transformation and the high-frequency sub-bands were interpolated. At last, the super-resolution image was obtained by S ＋ P inverse transformation. Experiments show that the image processed by this method has more detail information, better visual effect and higher peak signal noise ratio （PSNR）. Therefore, this algorithm is an effective method of super-resolution image reconstruction.