文件列表:

图像去噪C源代码\datatype.c (3783, 1999-05-12)
图像去噪C源代码\denoise.c (2327, 1999-05-12)
图像去噪C源代码\dwt.c (4162, 1999-05-12)
图像去噪C源代码\em.c (15914, 1999-05-12)
图像去噪C源代码\exvar.h (125, 1999-05-12)
图像去噪C源代码\generate.m (657, 1999-05-14)
图像去噪C源代码\hmt.h (1835, 1999-05-12)
图像去噪C源代码\lena (262144, 1999-05-12)
图像去噪C源代码\makefile (513, 1999-05-12)
图像去噪C源代码\post.c (9940, 1999-05-12)
图像去噪C源代码\psnr.m (88, 1999-05-12)
图像去噪C源代码\routines.c (5118, 1999-05-12)
图像去噪C源代码\rwmatrix.c (958, 1999-05-12)
图像去噪C源代码\showresult.m (981, 1999-05-14)
图像去噪C源代码\uhmt.c (1747, 1999-05-12)
图像去噪C源代码 (0, 2005-10-05)

This is a self-contained package including the 2-D DWT and inverse DWT routines. The included matlab routines are used to generate noisy example images and disply the result of denoising. INSTALLATION: download whmtc.tar "tar -xf whmtc.tar" decomposes the tar file into individual files. "make" on unix prompt to compile the c-codes. (or use other c compilers to compile the codes.) USAGE: denoise noisy_image output_image [tolerance] [size] command line arguments: noisy_image : file of input image to be denoised output_image: denoised image output filename [tolerance] : convergence error criterion for HMT training Default = 1e-3. If tolerance < 0, uHMT is used. [size] : size of image (default = 512) EXAMPLES: Step 1: Run "generate.m" under matlab. "generate('lena','lena25',25.5);" -A noisy Lena image is generated and stored in "lena25". Step 2: Run "denoise lena25 lena25d 1e-3 512" under unix prompt. -Train wavelet HMT model for the noisy image. Then, use empirical Wiener filter to denoise the image, saving the result in "lena25d". Step 3: Run "showresult.m" under matlab to display the results. "showresult('lena','lena25','lena25d');" -Displays denoising results as well as PSNR values. Note: The following commands are equivalent in Step 2. denoise lena25 lena25d 1e-3 512 denoise lena25 lena25d 1e-3 denoise lena25 lena25d Step 1: Same as above. Step 2: Run "denoise lena25 lena25d -1" under unix prompt. -Build a universal HMT (uHMT) and denoise the image. Step 3: Same as above. Notes: 1. Daubechies length 8 orthonormal wavelet is used as the wavelet transform. Other wavelets can be used by modifying the function "dwt2daub8()" in dwt.c. 2. The image input and output files are raw binary bit streams of the image matrix of "double" type in C. 3. While the wavelet HMT is being trained, the convergence error values are displyed. Related publications (postscript files available at www.dsp.rice.edu): -Basic description of wavelet HMT models: M. S. Crouse, R. D. Nowak, and R. G. Baraniuk, Wavelet-Based Signal Processing Using Hidden Markov Models, IEEE Transactions on Signal Processing (Special Issue on Wavelets and Filterbanks), April 19***. - 2-D wavelet HMT models and applications: H. Choi and R. G. Baraniuk, Multiscale texture segmentation using wavelet-domain hidden Markov models, Proc. 32nd Asilomar Conference, November 19***. H. Choi and R. G. Baraniuk, Image Segmentation using Wavelet-domain Classification, Proc. SPIE, Denver, July 1999. - 2-D wavelet HMT models and universal HMT (uHMT) models: J. K. Romberg, H. Choi, and R. G. Baraniuk, Bayesian Tree-Structured Image Modeling using Wavelet-domain Hidden Markov Models, Proc. SPIE, Denver, July 1999.

近期下载者：

相关文件：

评论：[我要评论] [举报此文件]

收藏者：