原理:主成分拉伸,(使主成分上的数据方差相等,进行高斯直方图匹配,再投影到原坐标)
题目:(操作比较渣,就用matlab+excel捯饬出的)
涉及函数
cova = cov(A) %求矩阵A的协方差矩阵,A的行表示样品,列表示特征
[coeff,latent,explained] = pcacov(cova) %已知协方差矩阵cova,求特征向量coeff,特征值latent,特征值(主成分)所占方差比例explained
[V,D] = eig(cova) %求协方差矩阵cova的特征向量V和特征值D
6.11
Perform a principal components transformation of the data shown in Fig. 4.21 and then produce a simple linear contrast stretch on each of the components separately. Compare the result to that from Prob. 4.3.
covA=
1.5088 1.1452
1.1452 1.3247
EVectors=
0.7349 -0.6782
0.6782 0.7349
EValues =
[2.5657 0.2678]
CPercentage =
90.5482
9.4518
Matlab 计算两个主成分的亮度范围(原坐标轴范围1~10)
>>comp1Range = [1 1; 10 10]*[0.7349; 0.6782]
comp1Range =
1.4131
14.1310
Comp1Mean = 7.77205
>> comp2Range = [10 1; 1 10]*[-0.6782;0.7349]
Comp2Range =
-6.0471
6.6708
Comp2Mean = 0.31185
高斯直方图匹配,假设匹配亮度至太分布三个标准差范围占99.73%,则根据
距离均值三个标准差的正态分布的分位数分别为-2.99,2.99
设图像的最大值为Xmax,最小值为Xmin,均值mean = (Xmax+Xmin)/2
因为Ф((Xmax-mean)/sd)= Ф(2.99)
算得两个主成分的方差为2.13
主成分变换
x | y | c1 | c2 | 像素数 |
3 | 3 | 4.2393 | 0.1701 | 1 |
3 | 4 | 4.9175 | 0.905 | 2 |
4 | 3 | 4.9742 | -0.5081 | 1 |
4 | 4 | 5.6524 | 0.2268 | 5 |
4 | 5 | 6.3306 | 0.9617 | 4 |
5 | 4 | 6.3873 | -0.4514 | 3 |
5 | 5 | 7.0655 | 0.2835 | 8 |
5 | 6 | 7.7437 | 1.0184 | 3 |
6 | 5 | 7.8004 | -0.3947 | 5 |
6 | 6 | 8.4786 | 0.3402 | 6 |
6 | 7 | 9.1568 | 1.0751 | 2 |
7 | 6 | 9.2135 | -0.338 | 3 |
7 | 7 | 9.8917 | 0.3969 | 1 |
7 | 8 | 10.5699 | 1.1318 | 1 |
8 | 7 | 10.6266 | -0.2813 | 1 |
8 | 8 | 11.3048 | 0.4536 | 1 |
直方图匹配方法
第一主成分直方图匹配,用到excel统计函数NORM.INV
次序 | c1 | 像素数 | 累加 | 占比 | 正太分布分位数 |
1 | 4.2393 | 1 | 1 | 0.021276596 | 3.452262721 |
2 | 4.9175 | 2 | 3 | 0.063829787 | 4.527215773 |
3 | 4.9742 | 1 | 4 | 0.085106383 | 4.850711393 |
4 | 5.6524 | 5 | 9 | 0.191489362 | 5.913793173 |
5 | 6.3306 | 4 | 13 | 0.276595745 | 6.5089929 |
6 | 6.3873 | 3 | 16 | 0.340425532 | 6.895976625 |
7 | 7.0655 | 8 | 24 | 0.510638298 | 7.828855863 |
8 | 7.7437 | 3 | 27 | 0.574468085 | 8.171981418 |
9 | 7.8004 | 5 | 32 | 0.680851064 | 8.773320371 |
10 | 8.4786 | 6 | 38 | 0.808510638 | 9.630306827 |
11 | 9.1568 | 2 | 40 | 0.85106383 | 9.989394722 |
12 | 9.2135 | 3 | 43 | 0.914893617 | 10.69338861 |
13 | 9.8917 | 1 | 44 | 0.936170213 | 11.01688423 |
14 | 10.5699 | 1 | 45 | 0.957446809 | 11.43947661 |
15 | 10.6266 | 1 | 46 | 0.978723404 | 12.09183728 |
16 | 11.3048 | 1 | 47 | 1 | #NUM! |
第二主成分直方图匹配
次序 | c2 | 像素数 | 累加 | 占比 | 正态分布分位数 |
3 | -0.5081 | 1 | 1 | 0.021276596 | -4.007937279 |
6 | -0.4514 | 3 | 4 | 0.085106383 | -2.609488607 |
9 | -0.3947 | 5 | 9 | 0.191489362 | -1.546406827 |
12 | -0.338 | 3 | 12 | 0.255319149 | -1.089357985 |
15 | -0.2813 | 1 | 13 | 0.276595745 | -0.9512071 |
1 | 0.1701 | 1 | 14 | 0.29787234 | -0.818178389 |
4 | 0.2268 | 5 | 19 | 0.404255319 | -0.204350924 |
7 | 0.2835 | 8 | 27 | 0.574468085 | 0.711781418 |
10 | 0.3402 | 6 | 33 | 0.70212766 | 1.441878389 |
13 | 0.3969 | 1 | 34 | 0.723404255 | 1.5749071 |
16 | 0.4536 | 1 | 35 | 0.744680851 | 1.713057985 |
2 | 0.905 | 2 | 37 | 0.787234043 | 2.009163374 |
5 | 0.9617 | 4 | 41 | 0.872340426 | 2.734776877 |
8 | 1.0184 | 3 | 44 | 0.936170213 | 3.556684227 |
11 | 1.0751 | 2 | 46 | 0.978723404 | 4.631637279 |
14 | 1.1318 | 1 | 47 | 1 | #NUM! |
这里我将#NUM!分别赋予主成分范围的最大值(也不知道可行不可行)
拉伸后数据 | 投影至原坐标轴 | 取整 | 像素数 | |||
c1' | c2' | x' | y' | x | y | |
3.4523 | 0.8182 | 3.0920 | 1.7400 | 3 | 2 | 1 |
4.5272 | 2.0092 | 1.9644 | 4.5469 | 2 | 5 | 2 |
4.8507 | 4.0079 | 6.2830 | 0.3443 | 6 | 1 | 1 |
5.9138 | 0.2044 | 4.4846 | 3.8606 | 4 | 4 | 5 |
6.5090 | 2.7348 | 2.9287 | 6.4242 | 3 | 7 | 4 |
6.8960 | 2.6095 | 6.8376 | 2.7591 | 7 | 3 | 3 |
7.8289 | 0.7118 | 5.2707 | 5.8326 | 5 | 6 | 8 |
8.1720 | 3.5567 | 3.5934 | 8.1560 | 4 | 8 | 3 |
8.7733 | 1.5464 | 7.4963 | 4.8136 | 7 | 5 | 5 |
9.6303 | 1.4419 | 6.0994 | 7.5909 | 6 | 8 | 6 |
9.9894 | 4.6316 | 4.2000 | 10.1786 | 4 | 10 | 2 |
10.6934 | 1.0894 | 8.5974 | 6.4517 | 9 | 6 | 3 |
11.0169 | 1.5749 | 7.0282 | 8.6291 | 7 | 9 | 1 |
11.4395 | 6.6708 | 3.8827 | 12.6606 | 4 | 10 | 1 |
12.0918 | 0.9512 | 9.5314 | 7.5016 | 10 | 8 | 1 |
14.1310 | 1.7131 | 9.2231 | 10.8426 | 9 | 10 | 1 |
10 |
|
|
| 3 |
|
|
|
| 1 |
|
9 |
|
|
|
|
|
| 1 |
|
|
|
8 |
|
|
| 3 |
| 6 |
|
|
| 1 |
7 |
|
| 4 |
|
|
|
|
|
|
|
6 |
|
|
|
| 8 |
|
|
| 3 |
|
5 |
| 2 |
|
|
|
| 5 |
|
|
|
4 |
|
|
| 5 |
|
|
|
|
|
|
3 |
|
|
|
|
|
| 3 |
|
|
|
2 |
|
| 1 |
|
|
|
|
|
|
|
1 |
|
|
|
|
| 1 |
|
|
|
|
| 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 |