SIFT_Matlab

所属分类:图形图像处理
开发工具:matlab
文件大小:1228KB
下载次数:1327
上传日期:2008-04-01 14:10:36
上 传 者gsammi
说明:  一个人写的sift代码,matlab实现,比起原作者的效果差些,但是也不错。(原作者没有公开源码)

文件列表:
SIFT_YantaoNoemie\demo-data\image068.JPG (14060, 2003-02-17)
SIFT_YantaoNoemie\demo-data\image069.JPG (13579, 2003-02-17)
SIFT_YantaoNoemie\demo-data\object0024.view01.png (578897, 2003-04-25)
SIFT_YantaoNoemie\demo-data\object0024.view03.png (574557, 2003-04-25)
SIFT_YantaoNoemie\demo-data\Thumbs.db (71680, 2006-11-20)
SIFT_YantaoNoemie\demo-data (0, 2006-11-21)
SIFT_YantaoNoemie\descriptor\do_descriptor.m (5290, 2006-11-17)
SIFT_YantaoNoemie\descriptor (0, 2006-11-20)
SIFT_YantaoNoemie\do_demo_1.m (867, 2006-11-20)
SIFT_YantaoNoemie\do_demo_2.asv (2126, 2006-11-20)
SIFT_YantaoNoemie\do_demo_2.m (1189, 2006-11-20)
SIFT_YantaoNoemie\do_demo_3.m (1171, 2006-11-20)
SIFT_YantaoNoemie\do_demo_4.m (1186, 2006-11-20)
SIFT_YantaoNoemie\do_sift.m (3972, 2006-11-20)
SIFT_YantaoNoemie\key-location\do_extrefine.m (3799, 2006-11-15)
SIFT_YantaoNoemie\key-location\do_localmax.asv (1992, 2006-11-17)
SIFT_YantaoNoemie\key-location\do_localmax.m (1991, 2006-11-17)
SIFT_YantaoNoemie\key-location (0, 2006-11-17)
SIFT_YantaoNoemie\main.m (140, 2006-05-04)
SIFT_YantaoNoemie\match\do_match.asv (4748, 2006-11-15)
SIFT_YantaoNoemie\match\do_match.m (4879, 2006-11-20)
SIFT_YantaoNoemie\match (0, 2006-11-17)
SIFT_YantaoNoemie\orientation\do_orientation.m (2366, 2006-11-17)
SIFT_YantaoNoemie\orientation (0, 2006-11-17)
SIFT_YantaoNoemie\scale-space\do_diffofg.m (464, 2006-11-14)
SIFT_YantaoNoemie\scale-space\do_gaussian.m (2639, 2006-11-14)
SIFT_YantaoNoemie\scale-space\smooth.m (215, 2006-11-14)
SIFT_YantaoNoemie\scale-space (0, 2006-11-17)
SIFT_YantaoNoemie\sift_demo.m (1637, 2006-11-14)
SIFT_YantaoNoemie\util\appendimages.m (359, 2006-11-15)
SIFT_YantaoNoemie\util\imreadbw.m (301, 2006-11-16)
SIFT_YantaoNoemie\util\plotsiftdescriptor.asv (3408, 2006-11-17)
SIFT_YantaoNoemie\util\plotsiftframe.m (1853, 2006-11-20)
SIFT_YantaoNoemie\util\plotss.m (640, 2006-11-20)
SIFT_YantaoNoemie\util\tightsubplot.m (1859, 2006-11-20)
SIFT_YantaoNoemie\util (0, 2006-11-20)
SIFT_YantaoNoemie (0, 2006-11-20)

SCALE INVARIANT FEATURE TRANSFORM IMPLEMENTATION YANTAO ZHENG, NOEMIE PHULPIN yantaozheng@gmail.com, S0600506@nus.edu.sg This is a MATLAB implementation of the SIFT keypoint detector and descriptor [1]. FUNCTIONS The implementation consists of the following functions. They are organized in the folders based on the functionality. These functions are self-contained and can be utilized independently. do_gaussian: generate Gaussian scale space of input image do_diffofg: generate Difference of Gaussian (DoG) scale space do_localmax: select local extrema as the potential keypoints do_extrefine: refine the keypoints by discarding the ones with low contrast and along an edge do_orientation: compute the orientation of a support region of keypoint do_descriptor: compute the descriptor of a keypoint based on image gradients. do_match: match two images based on the nearest neighbor principle and spatial consistency. do_sift: generate the SIFT descriptors for a given input image. It basically executes all the functions above. RUN THE PROGRAM If the purpose of using this software is to compute SIFT descriptors of images, users only need to handle do_sift function. do_sift takes an image as input and generate keypoints and descriptors as output. Detailed structures about input/output can be found in the program header comments. There are several demo programs avaiable. They are the examples to run the SIFT programs. Execute the demos as below: From MATLAB prompt > do_demo_1 IMPORTANT PARAMETERS The performance of SIFT image matching depends on several parameters. In do_sift function, tuning the following parameters can achive different number of SIFT keypoints and descriptors; and therefore, different matching performance. The values are the default ones given by Lowe. S=3 ; Number of sub-levels per octave omin= -1 ; Starting octave number O = 4; Max octave level thresh = 0.04 / S / 2 ; Contrast response threshold r = 15 ; Edge response threshold NBP = 4 ; Number of spatial bins NBO = 8 ; Number of orientation bins [1] D. G. Lowe, "Distinctive image features from scale-invariant keypoints," IJCV, vol. 2, no. 60, pp. 91 110, 2004.

近期下载者

相关文件


收藏者