In recent years, the study of image super-resolution reconstruction technology has been paid much attention to, because it can improve image recognition accuracy and recognition ability. One of the difficult problems is how to ensure the reconstruction quality of image edge texture area. In this paper, a single image super-resolution reconstruction approach based on wavelet domain is proposed. Firstly, the non-subsampled wavelet transform (NSWT) is applied to the input image, according to the multi-directionality of wavelet transform, three kinds of multi-angle templates are proposed, and each subband contour is estimated by total variation model (TV model) to determine its optimal direction. Then, the multi-angle templates and bicubic B-spline interpolation are used to interpolate the subbands. Finally, the non-subsampled wavelet inverse transform is implemented. This approach makes edge information and texture information of the reconstructed images more precise, and overcomes some deficiencies such as edge blurring, edge serration, as well as distortion of texture region, caused by traditional interpolation reconstruction approaches, such as bilinear interpolation and bicubic interpolation, etc. The quality of reconstructed image is improved. This approach can be used in image monitoring, remote sensing image analysis, medical image processing, and so on. A large number of simulation experiments verify the effectiveness of the proposed approach.
Key words
non-subsampled wavelet transform /
contour template /
variational calculus model /
directional interpolation /
super-resolution reconstruction /
edge /
high-frequency subbands /
multi-angle /
template matrix
{{custom_keyword}} /
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}
References
[1] Kim S P,Bose N K,Valenzuela H M.Recursive reconstruction of high resolution image from noisy undersampled multiframes[J].Acoustics Speech & Signal Processing IEEE Transactions on,1990,38(6):1013-1027.
[2] 卫保国,惠卫华.自适应耦合方向扩散和冲击滤波的图像放大[J].中国图象图形学报,2011,16(4):533-537.
[3] Zhang L,Wu X.An edge-guided image interpolation algorithm via directional filtering and data fusion[J].IEEE Transactions on Image Processing,2006,15(8):2226-38.
[4] Demirel H,Anbarjafari G.Image resolution enhancement by using discrete and stationary wavelet decomposition[J].IEEE Transactions on Image Processing,2011,20(5):1458-60.
[5] Mueller N,Lu Y,Do M N.Image interpolation using multiscale geometric representations[J].Proc SPIE,2007,13(6498):341-353.
[6] Getreuer P.Image Interpolation with contour stencils[J].Quality of Life Research,2011,1(4):389-399.
[7] Getreuer P.Contour Stencils:Total variation along curves for adaptive image interpolation[J].Siam Journal on Imaging Sciences,2011,4(3):954-979.
[8] 王相海,毕晓昀,傅博,等.多方向模板变分模型的单幅图像超分辨率重建[J].中国图象图形学报,2016,21(8):977-985.
[9] Rudin L,Osher S,Fatemi E.Nonlinear total variation based noise removal algorithms[J].Physica D:Nonlinear Phenomena,1992,60(1-4):259-268.
[10] Dong C,Chen C L,He K,et al.Learning a Deep Convolutional Network for Image Super-Resolution[M].Switzerland:Springer International Publishing,2014.184-199.
{{custom_fnGroup.title_en}}
Footnotes
{{custom_fn.content}}
Funding
National Natural Science Foundation of China (No.1671439, No.61402214); Innovation Team Support Plan of Colleges and Universities in Liaoning Province (No.LT2017013)
{{custom_fund}}