KB-Detector-src
所属分类：matlab编程
开发工具：matlab
文件大小：244KB
下载次数：113
上传日期：2011-09-15 10:03:24
上 传 者：
湘大
说明： 功能说明：输入一幅图像，能对此图像检测关键点，这些点通常对应图像中的重要对象，可用于对象识别，也可用来作为特征进行图像分类。运行其中的例程Ldx_GoSalScale.m，就可以知道此程序的用法（调试环境matlab7.0）。
(Function: input image, this image can detect critical points, these points usually correspond to an important object of the image can be used for object recognition, can also be used as features for image classification. Run one of routine Ldx_GoSalScale.m, you can know the usage of this program (debugging environment matlab7.0).)
文件列表:
KB-Detector-src\CalcEntropyScalesOpt.c (33353, 2005-09-21)
KB-Detector-src\CalcScaleSaliency.dll (28672, 2005-09-21)
KB-Detector-src\drawcircle.m (259, 2009-03-10)
KB-Detector-src\General.h (336, 2005-09-21)
KB-Detector-src\GoSalScale.m (914, 2009-03-10)
KB-Detector-src\GreedyCluster.m (601, 2005-09-21)
KB-Detector-src\Hists.c (1306, 2005-09-21)
KB-Detector-src\Hists.h (746, 2005-09-21)
KB-Detector-src\image_0001.jpg (10377, 2005-09-21)
KB-Detector-src\image_0001_out.jpg (121934, 2005-09-21)
KB-Detector-src\Ldx_GoSalScale.asv (952, 2011-03-08)
KB-Detector-src\Ldx_GoSalScale.m (1073, 2011-03-27)
KB-Detector-src\Makefile (725, 2005-09-21)
KB-Detector-src\ScaleSaliency.dll (225280, 2003-10-01)
KB-Detector-src (0, 2011-09-15)
Kadir/Brady Feature Detector (Scale Saliency) Version 1.5
For non commercial purposes only.
Copyright Timor Kadir 19***-2004.
To compile use the Makefile or type
mex -O -DFASTPLOG CalcEntropyScalesOpt.c Hists.c -output CalcScaleSaliency
For an example of how to use the binaries run GoSalScale at the matlab prompt. The output should be the same as image_0001_out.jpg.
1) CalcScaleSaliency
This function finds the salient points in an image.
Usage:
Y=CalcScaleSaliency(uint8(double(Im)./div),StartScale,StopScale, Nbins, Sigma,AA);
Im : input image (can be 8-bit grey-level, 2D 8-bit e.g. UV, 3D 8-bit e.g. RGB)
Nbins : number of bins for histogram
div : 256/NBINS
Sigma : sigma of kernel for Parzen estimator (set NBINS to 0)
StartScale : Minimum scale
StopScale : Maximum scale
AA : Use anti-aliased sampling for generating the histograms (see Kadir, Zisserman and Brady ECCV 2004).
Scale means the radius of the circle. The function expects the image to be quantised to the number of bins specified. To do that we just divide by 256/nbins.
The output, Y, is a list of points in the following format:
X Position
Y Position
Scale
Entropy (HD)
Inter-scale saliency (WD)
Scale Saliency (YD)
2) GreedyCluster
Usage :
C=GreedyCluster(Y,wt,yt)
Y : List of salient features from CalcScaleSaliency
wt : Threshold on inter-scale saliency (WD)
yt : Threshold on saliency (YD)
This, as its name implies, is a greedy clusterer which starts with the highest saliency feature and works down removing any features that are too `close' in x,y,s. Here, close means within the support of the current feature i.e its diameter. Its written in matlab so can be modified easily.
GreedyCluster has two parameters : Yt a global saliency threshold and Wt a global inter-scale saliency threshold (set between 0 and 2). Setting the later high forces the selection of features that are more scale-localised or isotropic.
The output, C, is a list of clustered salient regions in the same format as above.
近期下载者：
相关文件：
收藏者：