RH
所属分类:图形图像处理
开发工具:Visual C++
文件大小:114KB
下载次数:54
上传日期:2009-05-19 00:21:29
上 传 者:
bluewhale520
说明: 图像特征提取与识别算法,自己编的,大家用下感觉如何?
(Image Feature Recognition)
文件列表:
IMAGES (0, 2001-06-12)
IMAGES\breast_cell.pbm (51570, 1998-02-24)
IMAGES\eggs.pbm (40901, 1998-02-24)
IMAGES\plates.pbm (46923, 1998-02-24)
Makefile (734, 1998-03-04)
include (0, 2001-06-12)
include\globals.h (1256, 1998-03-01)
include\hough.h (14316, 1998-03-03)
include\rht.h (990, 1998-02-24)
src (0, 2001-06-12)
src\Xdraw_func.c (85066, 1998-02-24)
src\geo_sym_ellipse.c (53519, 1998-02-24)
src\image.c (17686, 1998-03-01)
src\init_X_stuff.c (66883, 1998-03-01)
src\init_hough.c (1335, 1998-02-24)
src\lin_algebra.c (46146, 1998-03-01)
src\line.c (2924, 1998-02-24)
src\menu_callback_functions.c (12956, 1998-03-01)
src\misc.c (2358, 1998-02-24)
src\node.c (13639, 1998-03-01)
src\noise.c (1831, 1998-02-24)
src\pht_ellipse.c (39741, 1998-03-01)
src\point.c (7705, 1998-02-24)
src\recog.c (15549, 1998-03-03)
src\rht.c (55106, 1998-02-24)
src\rht_data_structures.c (8534, 1998-02-24)
src\sht_ellipse.c (43070, 1998-03-01)
src\sht_line.c (19393, 1998-02-24)
Program name:
-------------
recog
Author:
-------
Robert McLaughlin (ram@ee.uwa.edu.au)
How to compile:
---------------
type:
make
If compiling on a machine running Linux or SunOS,
uncomment the appropriate lines in the Makefile.
By default, the program uses the compiler gcc.
To change this to the compiler cc, read the Makefile.
The program was written on a Unix machine in ANSI C,
and should compile on a Unix machine which has X11 installed.
The program was not written for Windows 95 machines.
The program has been tested on SGI and Linux machines using both
cc and gcc compilers.
It has been tested on SunOS machines using the gcc compiler.
Usage:
------
recog - begin with a blank image
recog filename.pbm - load an edge-detected pbm image.
Edges should be approx. 2 pixels thick.
e.g. recog IMAGES/eggs.pbm
This program has been tested with pbm images generated by xv.
The xv program is available from:
http://www.trilon.com/xv/
Note that the Standard Hough Transform and Probabilistic Hough
Transform for Ellipse detection are VERY SLOW algorithms.
I strongly recommend that the first time you use these algorithms,
just draw a single ellipse on the screen and then run the algorithm.
Then try it with two ellipses. The computation time required for
the algorithms is roughly related to the square of the number of
black pixels on the screen.
All algorithms degrade rapidly when subject to speckle noise.
Also, computation times tend to increase rapidly under speckle
noise. If you choose to add speckle noise to an image, I strongly
recommend that you begin with a value like 0.5 or 1.
To change the size of the default window (which is 750 x 550)
you need to change the #define statements for
WINDOW_SIZE_X and WINDOW_SIZE_Y in the file
include/hough.h
Description
-----------
The program allows the user to hand draw simple shapes
using the mouse. The program will then perform a recognition
upon the image.
Four ellipse detection algorithms have been implemented:
1. Randomized Hough Transform:
Details have been submitted in a paper to
Pattern Recognition Letters:
"Randomized Hough Transform: Improved Ellipse Detection with
Comparison", R. A. McLaughlin, Pattern Recognition Letters, 19***.
For details, check out the web site:
http://ciips.ee.uwa.edu.au/Papers/Journal_Papers/19***/01/Index.html
2. Standard Hough Transform
A standard ellipse detection method which makes use of the
Hough Transform.
This is an implementation of the algorithm described in the paper:
"Detecting Partially Occluded Ellipses Using the Hough
Transform", H. K. Yuen, J. Illingworth and J. Kittler,
Image and Vision Computing, Vol. 7, No. 1, 1***9, pp 31-37.
3. Probabilistic Hough Transform
A probabilistic implementation of the above standard algorithm
4. Geometric Symmetry
An implementation of the method detailed in the paper:
"A Fast Ellipse/Circle Detector Using Geometric Symmetry",
Chun-Ta Ho and Ling-Hwei Chen, Pattern Recognition,
Vol. 28, No. 1, 1995, pp 117-124.
A brief explanation of each algorithm can be found in the paper:
"Randomized Hough Transform: Improved Ellipse Detection with
Comparison", R. A. McLaughlin, Pattern Recognition Letters, 19***.
http://ciips.ee.uwa.edu.au/Papers/Journal_Papers/19***/01/Index.html
IMPORTANT NOTE:
---------------
Implementations of the Standard Hough Transform and the
Probabilistic Hough Transform have been improved since
the journal paper was written.
Thus both algorithms will achieve slightly better results with
the real-world images than was published in the paper.
The implementation of the Randomized Hough Transform
may give different results when run several times on
the same image.
This can be seen if you process the test image
IMAGE/breast_cell.pbm several times.
This is a side-effect of the tree structure used to
store points in the Hough parameter space.
The program assumes that edges in images are approximately
2 pixels thick.
Bugs and Disclaimer
-------------------
The documentation is not fully detailed and some minors bugs remain.
No support is provided for users to use or modify this software.
We can not and will not provide any technical support for
using this software.
Licence Information
-------------------
This software may be freely used for research and scholarly purposes.
No part of this software may be used as part of a commercial
software product without obtaining the prior written consent of the
author, Robert McLaughlin.
Release Dates:
--------------
Version 1.0 18 August, 1996
Version 1.1 25 May, 1997 Bug fixes
Version 1.2 18 August, 1997 Bug fixes
Version 1.3 24 January, 19*** Organised code into src, include
and IMAGES subdirectories.
Simplified Makefile.
Bug fixes.
Version 1.4 4 March, 19*** Implement code to handle displays
with different color depths.
Robert McLaughlin
4 March, 19***
近期下载者:
相关文件:
收藏者: