Matlab-vision包学习-Feature Detection,Extraction and Matching-提取描述子

这篇介绍提取Harris角点描述子的函数

函数/Function

函数名称：extractFeatures

功能：提取利用诸如BRISK/Harris/FREAK/SURF/Block等方法检测到的特征点的描述子

语法：[features,validPoints] = extractFeatures(I,points);

            [features,validPoints] = extractFeatures(I,points,Name,Value); 

其中，I为2-D灰度图像，points为特征点检测算法获取的特征点(可能为BRISKPoints/CornerPoints/MSERRegions/SURFPoints对象或Mx2的[x,y]坐标矩阵)，Name为用一对单引号包含的字符串，Value为对应Name的值

Name&Value参数

Name	Value
’Method'	默认为‘Auto'，可以为'BRISK'/'FREAK'/'SURF'/'Block'/'Auto'，具体见“Method及其含义”表
'BlockSize'	对应’Block'方法的邻域正方形的大小
'SURFSize'	获取的SURF描述子的长度，默认值为64，还可以设置为128，设置为128在提高精度的同时增加了匹配的时间

Method&及其含义

Method	含义
'BRISK'	对应输入为BRISKPoints对象，函数将输出变量’validPoints'对象中的方向(Orientation)属性设置为直径的方向
‘FREAK'	Fast Retina KeyPoint(FREAK)，对应输入为CornerPoints对象，函数将输出变量’validPoints'对象中的方向(Orientation)属性设置为直径的方向
’SURF'	对应输入为SURFPoints/MSERRegions对象，当输入为SURFPoints，函数将输出变量’validPoints'对象中的方向(Orientation)属性设置为直径的方向；当输入为MSERRegions（需要理解，后续添加）
'Block'	简单正方形邻域，仅对输入为Mx2的[x,y]坐标矩阵时可用。‘Block’方法只会抽取正方形领域中的所有点在图像内的特征，所以，输出的点可能会比输入的点少！
‘Auto'	函数的默认值，对应不同的输入对象，调用不同的方法

举例：

<pre name="code" class="plain">close all;
clear all;
clc;
% Extract corner features from an image.
I = imread('cameraman.tif');
corners = detectHarrisFeatures(I);
[features, valid_corners] = extractFeatures(I, corners);

figure; imshow(I); hold on
% Plot valid corner points
plot(valid_corners);
title('Harris')

% Extract SURF features
I = imread('cameraman.tif');
points = detectSURFFeatures(I);

[features, valid_points] = extractFeatures(I, points);

% Visualize 10 strongest SURF features, including their scale, and
% orientation which was determined during the descriptor extraction
% process.
figure; imshow(I); hold on;
plot(valid_points.selectStrongest(10),'showOrientation',true);
title('SURF')

% Extract MSER features
I = imread('cameraman.tif');
regions = detectMSERFeatures(I);

% Use MSER with SURF feature descriptor
[features, valid_points] = extractFeatures(I, regions);

% Visualize SURF features corresponding to the MSER ellipse centers
% along with scale and orientation.
figure; imshow(I); hold on;
plot(valid_points,'showOrientation',true);
title('MSER')