TurboPixels
所属分类:图形图像处理
开发工具:matlab
文件大小:95KB
下载次数:96
上传日期:2013-12-05 10:04:09
上 传 者:
aaCherish
说明: 图像预分割为一个个超像素,在用谱聚类或者其他方法做图像分割时减轻计算量
(Image pre-segmented into one super-pixel, reducing the amount of computation using spectral clustering or any other way to do image segmentation)
文件列表:
TurboPixels\clip.m (150, 2007-04-13)
TurboPixels\convolve.c (10499, 2007-04-13)
TurboPixels\convolve.h (963, 2007-04-13)
TurboPixels\corrDn.c (4560, 2007-04-13)
TurboPixels\demo_superpixels.m (175, 2007-04-13)
TurboPixels\display_contour.m (1210, 2007-04-13)
TurboPixels\display_logical.m (687, 2007-04-13)
TurboPixels\DT.c (3779, 2007-04-13)
TurboPixels\edges.c (22123, 2007-04-13)
TurboPixels\evolve_height_function.m (837, 2007-04-13)
TurboPixels\evolve_height_function_N.m (5620, 2007-04-13)
TurboPixels\get_binary_img_from_seeds.m (572, 2007-04-13)
TurboPixels\get_full_speed.c (1768, 2007-04-13)
TurboPixels\get_initial_seeds.m (2412, 2007-04-13)
TurboPixels\get_speed_based_on_boundaries.m (333, 2007-04-13)
TurboPixels\get_speed_based_on_gradient.m (2166, 2007-04-13)
TurboPixels\get_superpixel_boundaries.m (2153, 2007-04-13)
TurboPixels\hasToRecomputeBand.m (143, 2007-04-13)
TurboPixels\height_function_change_rate.m (295, 2007-04-13)
TurboPixels\height_function_der.c (1417, 2007-04-13)
TurboPixels\height_function_der.mexw32 (6656, 2009-10-15)
TurboPixels\height_function_grad.c (3836, 2007-04-13)
TurboPixels\lizard.jpg (37923, 2007-04-13)
TurboPixels\local_min.c (1259, 2007-04-13)
TurboPixels\lsmlib\computeDistanceFunction2d.c (4147, 2007-04-13)
TurboPixels\lsmlib\computeDistanceFunction2d.m (1736, 2007-04-13)
TurboPixels\lsmlib\computeExtensionFields2d.c (5633, 2007-04-13)
TurboPixels\lsmlib\computeExtensionFields2d.m (1908, 2007-04-13)
TurboPixels\lsmlib\doHomotopicThinning.c (3461, 2007-04-13)
TurboPixels\lsmlib\FMM_Callback_API.h (3614, 2007-04-13)
TurboPixels\lsmlib\FMM_Core.c (16672, 2007-04-13)
TurboPixels\lsmlib\FMM_Core.h (10332, 2007-04-13)
TurboPixels\lsmlib\FMM_Heap.c (12809, 2007-04-13)
TurboPixels\lsmlib\FMM_Heap.h (7062, 2007-04-13)
TurboPixels\lsmlib\lsm_fast_marching_method.h (14479, 2007-04-13)
TurboPixels\lsmlib\lsm_FMM_field_extension2d.c (38341, 2007-04-13)
TurboPixels\make.m (773, 2007-04-13)
TurboPixels\superpixels.m (1762, 2007-08-18)
TurboPixels\upConv.c (6369, 2007-04-13)
... ...
Instalation and running instructions for superpixel code
--------------------------------------------------------
DISCLAIMER:
This is a beta version of the superpixel code. It hasn't been
thoroughly tested on different architectures. It was tested on
a Linux system with Matlab 7.1, with gcc and g++ compilers.
The affinity function that is used in the code is simple and
is not sufficient to provide good results on all possible images.
The domain of the images affects the choice of the affinity
function.
This file contains instructions on running the superpixel code.
Instalation:
------------
- Put all the code into a directory (for example 'my_dir')
- Run 'cd my_dir'
- Start matlab
- The subdirectory 'lsmlib' needs to be added to matlab's path. This is done
in demo_superpixels.m, or you can do it manually by typing 'addpath('lsmlib')'
- Run 'make'. To work, MEX needs to be properly configured. Ignore the warnings.
- Load an image 'img = im2double(imread(image_filename))'. This ensures that
the loaded image is in double format with every component in 0..1 range. Another
method for loading images can be used as long as the resulting image is a 2D
array of doubles.
The image can be RGB or grayscale, although it only matters for displaying purposes.
- Run the superpixels() function in superpixels.m. The header describes the
different parameters.
- demo_superpixels.m gives one example of using the code.
Examples:
1) Getting 1500 superpixels without intermediate display
img = im2double(imread('lizard.jpg'));
[phi,boundary,disp_img] = superpixels(img, 1500);
imagesc(disp_img);
2) Getting superpixels with an intermediate display every 5 timesteps
img = im2double(imread('lizard.jpg'));
[phi,boundary,disp_img] = superpixels(img, 1500, 5);
imagesc(disp_img);
3) Getting superpixels while saving every frame of the evolution into a movie
img = im2double(imread('lizard.jpg'));
[phi,boundary,disp_img, frames] = superpixels(img, 1500, 1);
disp('Press any key to view the evolution movie');
pause;
movie(frames);
近期下载者:
相关文件:
收藏者: