实例介绍
【实例简介】
介绍了一种基于MAP的图像超分辨率重建算法
chanced image. In the iteration steps. the ‖/‖ (13) choice of a utilizes the information available at each iteration step in the enhancement process of the high- Where d is the iteration termination coefficient resolution image [11] o, The following properties of a is incorporated in 5. Experimental Results proposed algorithm a is proportional to y-Ax I In the experiments, we constructed high-resolution images from four low-resolution images. Firstly, each a is inversely proportional to Qx low-resolution image was first obtained by sub a is larger than zero sampling the original image, and then was o satisfy the above conditions, the proposed reconstructed back to their original dimensions so that regularization functional is designed as follows: quantitative comparisons could be made. In order to make a further demonstration about the performance of 0≈1M2∥y-Ax‖ our algorithm, the low-resolution images were Qxk‖2+r ted by 2%% gaussian d Where r is thc control parameter which prcvcnts thc reconstructed again. A image quality evaluation method of PsnR was used in this paper. PSNR is dominator from becoming zero./ is the modified inec factor of a 255*N 4.2. Iteration method PSNR=100g(文-x (14) Where N is the total number of pixels, and X and The gradient optimization technique was used to X are the reconstructed high-resolution image and the solve the high-resolution image, and it converges to a al image respectively. local minimum of the objective function of (9).A The original Lena image, one frame of the sub- sequence of iterates are generated by Sampled low-resolution images and one frame of low k1=Xk+[A(AA+anQ Q)xk(12) resolution images that was corrupted by 2% additive The criterion that is used to terminate the iteration is Gaussian noise are shown in Figure 2 defined as Figure 2.(a) Original Lena image. (b)One frame of low-resolution images (c)One frame of low-resolution images corrupted by 2%Gaussian noise The proposed algorithm and the conventional By By comparison, we can see that the high-resolution algorithm were used to reconstruct the high-resolution images reconstructed by the proposed algorithm almost images with the two series of low-resolution images have the same quality as the high-resolution espectively. In the conventional algorithm, the whose quality are the best of the conventional regularization parameter was chosen 1010 algorithm. Using thc conventional algorithm, we may 10 and 10. The psnr of the reconstructed high- need to try many times to find the best fixed resolution image for each case is presented in Table 1 regularized parameter. However, the proposed and the high-resolution image reconstructed by thc algorithm can choose and update the regularized proposed algorithm and the high-resolution image parameter automatically in the iterative process and can whose quality are the best of the images reconstructed also get best result by the conventional algorithm are shown in figure 3 Proceedings of the Third International Conference on Image and Graphics(ICIG 04) 7695-22440/04$20.00@2004IEEE COMPUTER SOCIETY Table 1. Psnr of the proposed reconstruction algorithm and the conventional reconstruction algorithm with fixed regularization parameter. a=10 a=10a=10-a=10-Proposed Algorithm No noise.415 35.317 36.549 34.538 36.201 2% noise29.6793102130324 27.856 30.776 (a) Figure 3. Experimental results: a The best one of the conventional algorithm from original low resolution images(b)The proposed algorithm from original low-resolution images (c) The best one of the conventional algorithm from corrupted low-resolution images. (d)The proposed algorithm from corrupted low-resolution images 6. Conclusions 5R. R. Schulz and R L. Stevenson, Extraction of high A super resolution reconstruction algorithm to resolution frames from video sequences, IEEE Tran estimate a high-resolution image from a set of sub Image Processing, vol. 5, Junc. 1996, pp. 996-1011 sampled low-resolution images has been proposed. The ed. the [6] B.C. Tom and A K. Katsaggelos, " Reconstruction of a proposed algorithm has been validated experimentally high-resolution image by simultaneous matching, Lena image frames. It has shown that the restoration, and interpolation of low-resolution images, Proc. 1995 IEEE Int. Conf. Image Processing, vol. 2 proposed algorithm not only can choose and update the Washington, DC, Oct 1995, pp 539-542 regularized parameter automatically in the iterations, [7] M. Irani and S Peleg, "Improving resolution by image but can get desirable high-resolution reconstruction matching, CyGIP: Graphical Models and Image Proc Images vol.53,May.1991,pp.231-239 [8]M. Elad and A. Feuer, "Superresolution restoration of an Reference: image sequence: adaptive filtering approach, IEET [1 R.Y. Tsai and T.s. Huang, Multiple frame image Trans. Image Processing, vol 8, Mar. 1999 pp. 387- 395 estoration and matching "in Advances in Computer Vision and Image Processing. Greenwich, CT: JAI [9] J Fryer and G. Scarmana, "Algorithm Development for Press Inc, 1984, pp. 317-339 the enhancement of Photogrammetric Digital Images to [2]J. J. Clark, M R. Palmer, and P D. Laurence,"A ImproveDemGenerationhttp://www.isprs.org/comm transformation method for the reconstruction of ission/proceedings/papers/paper002 functions from nonuniformly spaced samples, IEEE [10]SC. Park, M.K. Park, and M.G. Kang, " Super resolution image reconstruction: a technical overview Trans. AcouSt., Speech, Signal Processing, vol. ASSP 33,1985,pp.1151-1165 Signal Processing Magazine, IEEE, Vol. 20, Issue: 3 [3]I. Stark and P. Oskoui, "Iligh resolution image recovery May.2003,pp.21-36 from image-plane arrays, using convex projections, "J. [11] E. Lcc and M.G. Kang,"Regularized adaptive high Op.Soc.Am.A,vol.6,1989,pp.1715-1726 resolution image reconstruction considering inaccurate 14A. M. Tekalp, M.K. Ozkan, and M.I. Sezan,High subpixel matching, IEEE Trans. Image Processing resolution image reconstruction from lower-resolution vl.12,July,2003,pp.826-83 image sequences and space varying image restoration, [12]S. Borman and R. Stevenson,""Spatial Resolution in PrUc: IEEE Int. Cun/ AcousticS, Speech and Signal Enhancement of Low-Resolution Image sequences Processing (ICASSP), San Francisco, CA, vol 3, Mar http://decsai.ugres/vip/doctorado/pvd/borman98rev 1992,pp.169172. spatial-resolution-enhancement-review. pdf. Proceedings of the Third International Conference on Image and Graphics(ICIG 04) 7695-22440/04$20.00@2004IEEE COMPUTER SOCIETY 【实例截图】
【核心代码】
介绍了一种基于MAP的图像超分辨率重建算法
chanced image. In the iteration steps. the ‖/‖ (13) choice of a utilizes the information available at each iteration step in the enhancement process of the high- Where d is the iteration termination coefficient resolution image [11] o, The following properties of a is incorporated in 5. Experimental Results proposed algorithm a is proportional to y-Ax I In the experiments, we constructed high-resolution images from four low-resolution images. Firstly, each a is inversely proportional to Qx low-resolution image was first obtained by sub a is larger than zero sampling the original image, and then was o satisfy the above conditions, the proposed reconstructed back to their original dimensions so that regularization functional is designed as follows: quantitative comparisons could be made. In order to make a further demonstration about the performance of 0≈1M2∥y-Ax‖ our algorithm, the low-resolution images were Qxk‖2+r ted by 2%% gaussian d Where r is thc control parameter which prcvcnts thc reconstructed again. A image quality evaluation method of PsnR was used in this paper. PSNR is dominator from becoming zero./ is the modified inec factor of a 255*N 4.2. Iteration method PSNR=100g(文-x (14) Where N is the total number of pixels, and X and The gradient optimization technique was used to X are the reconstructed high-resolution image and the solve the high-resolution image, and it converges to a al image respectively. local minimum of the objective function of (9).A The original Lena image, one frame of the sub- sequence of iterates are generated by Sampled low-resolution images and one frame of low k1=Xk+[A(AA+anQ Q)xk(12) resolution images that was corrupted by 2% additive The criterion that is used to terminate the iteration is Gaussian noise are shown in Figure 2 defined as Figure 2.(a) Original Lena image. (b)One frame of low-resolution images (c)One frame of low-resolution images corrupted by 2%Gaussian noise The proposed algorithm and the conventional By By comparison, we can see that the high-resolution algorithm were used to reconstruct the high-resolution images reconstructed by the proposed algorithm almost images with the two series of low-resolution images have the same quality as the high-resolution espectively. In the conventional algorithm, the whose quality are the best of the conventional regularization parameter was chosen 1010 algorithm. Using thc conventional algorithm, we may 10 and 10. The psnr of the reconstructed high- need to try many times to find the best fixed resolution image for each case is presented in Table 1 regularized parameter. However, the proposed and the high-resolution image reconstructed by thc algorithm can choose and update the regularized proposed algorithm and the high-resolution image parameter automatically in the iterative process and can whose quality are the best of the images reconstructed also get best result by the conventional algorithm are shown in figure 3 Proceedings of the Third International Conference on Image and Graphics(ICIG 04) 7695-22440/04$20.00@2004IEEE COMPUTER SOCIETY Table 1. Psnr of the proposed reconstruction algorithm and the conventional reconstruction algorithm with fixed regularization parameter. a=10 a=10a=10-a=10-Proposed Algorithm No noise.415 35.317 36.549 34.538 36.201 2% noise29.6793102130324 27.856 30.776 (a) Figure 3. Experimental results: a The best one of the conventional algorithm from original low resolution images(b)The proposed algorithm from original low-resolution images (c) The best one of the conventional algorithm from corrupted low-resolution images. (d)The proposed algorithm from corrupted low-resolution images 6. Conclusions 5R. R. Schulz and R L. Stevenson, Extraction of high A super resolution reconstruction algorithm to resolution frames from video sequences, IEEE Tran estimate a high-resolution image from a set of sub Image Processing, vol. 5, Junc. 1996, pp. 996-1011 sampled low-resolution images has been proposed. The ed. the [6] B.C. Tom and A K. Katsaggelos, " Reconstruction of a proposed algorithm has been validated experimentally high-resolution image by simultaneous matching, Lena image frames. It has shown that the restoration, and interpolation of low-resolution images, Proc. 1995 IEEE Int. Conf. Image Processing, vol. 2 proposed algorithm not only can choose and update the Washington, DC, Oct 1995, pp 539-542 regularized parameter automatically in the iterations, [7] M. Irani and S Peleg, "Improving resolution by image but can get desirable high-resolution reconstruction matching, CyGIP: Graphical Models and Image Proc Images vol.53,May.1991,pp.231-239 [8]M. Elad and A. Feuer, "Superresolution restoration of an Reference: image sequence: adaptive filtering approach, IEET [1 R.Y. Tsai and T.s. Huang, Multiple frame image Trans. Image Processing, vol 8, Mar. 1999 pp. 387- 395 estoration and matching "in Advances in Computer Vision and Image Processing. Greenwich, CT: JAI [9] J Fryer and G. Scarmana, "Algorithm Development for Press Inc, 1984, pp. 317-339 the enhancement of Photogrammetric Digital Images to [2]J. J. Clark, M R. Palmer, and P D. Laurence,"A ImproveDemGenerationhttp://www.isprs.org/comm transformation method for the reconstruction of ission/proceedings/papers/paper002 functions from nonuniformly spaced samples, IEEE [10]SC. Park, M.K. Park, and M.G. Kang, " Super resolution image reconstruction: a technical overview Trans. AcouSt., Speech, Signal Processing, vol. ASSP 33,1985,pp.1151-1165 Signal Processing Magazine, IEEE, Vol. 20, Issue: 3 [3]I. Stark and P. Oskoui, "Iligh resolution image recovery May.2003,pp.21-36 from image-plane arrays, using convex projections, "J. [11] E. Lcc and M.G. Kang,"Regularized adaptive high Op.Soc.Am.A,vol.6,1989,pp.1715-1726 resolution image reconstruction considering inaccurate 14A. M. Tekalp, M.K. Ozkan, and M.I. Sezan,High subpixel matching, IEEE Trans. Image Processing resolution image reconstruction from lower-resolution vl.12,July,2003,pp.826-83 image sequences and space varying image restoration, [12]S. Borman and R. Stevenson,""Spatial Resolution in PrUc: IEEE Int. Cun/ AcousticS, Speech and Signal Enhancement of Low-Resolution Image sequences Processing (ICASSP), San Francisco, CA, vol 3, Mar http://decsai.ugres/vip/doctorado/pvd/borman98rev 1992,pp.169172. spatial-resolution-enhancement-review. pdf. Proceedings of the Third International Conference on Image and Graphics(ICIG 04) 7695-22440/04$20.00@2004IEEE COMPUTER SOCIETY 【实例截图】
【核心代码】
标签：
好例子网口号：伸出你的我的手 — 分享！
网友评论
小贴士
感谢您为本站写下的评论，您的评论对其它用户来说具有重要的参考价值，所以请认真填写。
- 类似“顶”、“沙发”之类没有营养的文字，对勤劳贡献的楼主来说是令人沮丧的反馈信息。
- 相信您也不想看到一排文字/表情墙，所以请不要反馈意义不大的重复字符，也请尽量不要纯表情的回复。
- 提问之前请再仔细看一遍楼主的说明，或许是您遗漏了。
- 请勿到处挖坑绊人、招贴广告。既占空间让人厌烦，又没人会搭理，于人于己都无利。
关于好例子网
本站旨在为广大IT学习爱好者提供一个非营利性互相学习交流分享平台。本站所有资源都可以被免费获取学习研究。本站资源来自网友分享，对搜索内容的合法性不具有预见性、识别性、控制性，仅供学习研究，请务必在下载后24小时内给予删除，不得用于其他任何用途，否则后果自负。基于互联网的特殊性，平台无法对用户传输的作品、信息、内容的权属或合法性、安全性、合规性、真实性、科学性、完整权、有效性等进行实质审查；无论平台是否已进行审查，用户均应自行承担因其传输的作品、信息、内容而可能或已经产生的侵权或权属纠纷等法律责任。本站所有资源不代表本站的观点或立场，基于网友分享，根据中国法律《信息网络传播权保护条例》第二十二与二十三条之规定，若资源存在侵权或相关问题请联系本站客服人员，点此联系我们。关于更多版权及免责申明参见 版权及免责申明
支持(0) 盖楼(回复)