基于Haar小波域自学习的图像超分辨率方法

Haar Wavelet Domain Self-training Based Image Super-resolution

图像超分辨率重构是指由低分辨率图像来获得高分辨率图像的过程.为了能够有效的重构出高分辨率图像,提出一种基于Haar小波域自学习的图像超分辨率重构算法.该算法将高分辨率图像通过Haar小波变换后得到的近似子块L与已知的低分辨率图像联系起来,然后通过Bp神经网络来自学习Haar小波变换细节子块之间相近的自相似性,从而预测出高分辨率图像通过Haar小波变换后的三个细节子块H,V和D.最后由逆Haar小波变换重构高分辨率图像.实验表明由该算法重构的高分辨图像有很好的视觉效果和峰值信噪比.

Image super-resolution refers to the reconstruction of a high resolution image from one or a set of blurred low resolution images.This paper only pays attention to the kind of reconstruction from one blurred low resolution image.Many methods have been developed about this kind of reconstruction,most of which are interpolation methods.In this paper we propose a new super-resolution method which base on the haar wavelet transform and back-propagation network.This new method takes advantage of the relationship between the low resolution image and the approximate subband of the high resolution image＇s haar wavelet decomposition and uses the relationship to compute the approximate subband from the low resolution image.Considering the self-similarity between the detail subbands of the haar wavelet decomposition,firstly the method trains the back-propagation network to approximate the self-similarity relationship and then uses the trained network to predict the detail subbands of the high resolution image＇s haar wavelet decomposition.Once the approximate and detail subbands of the high resolution image＇s haar wavelet decomposition are obtained,the high resolution image can be reconstructed.This paper uses the peak signal to noise ratio（PSNR） to compare the reconstructed image with the original image.And experiments show that the PSNR and visual effect of the high resolution image reconstructed with the proposed method is very good.