Let us consider the following confusion matrix:

	Class observed
Class expected	1	2	3	4	5	6	7	8	9	Unclassified
1	99.78	-	-	-	-	0.22	-	-	-	-
2	3.92	96.08	-	-	-	-	-	-	-	-
3	-	-	96.73	3.27	-	-	-	-	-	-
4	-	-	0.51	94.12	5.37	-	-	-	-	-
5	-	-	-	20.48	79.52	-	-	-	-	-
6	8.50	-	-	-	-	91.50	-	-	-	-
7	-	-	0.65	0.22	-	-	97.60	1.53	-	-
8	-	-	-	-	-	-	-	100.00	-	-
9	-	-	-	-	-	-	-	-	100.00	-

 If we consider the first line, we get:

After classification of the pixels of the image, we expect that all the pixels which belong to class 1 have been classified in class 1. But we observe that:

• 99.78% of the pixels have been classified in class 1,
•  0.22% of the pixels have been classified in class 6.

The sum of the observations of each line of the matrix is equal to 100% . Here, for the line considered, we have 99.78+ 0.22= 100.

 Consider now the 2 following lines:

	Class observed
Class expected	1	2	3	4	5	6	7	8	9	Unclassified
7	-	-	0.65	0.22	-	-	97.60	1.53	-	-
9	-	-	-	-	-	-	-	-	100.00	-

After classification, we observe that :

• whereas we expect that all pixels which belong to class 7 have been classified in class 7:
•   0.65% have been classified in class 3,
•   0.22% have been classified in class 4,
• 97.60% have been classified in class 7,
•   1.53% have been classified in class 8.

The sum of the percentages of the line is equal to 0.65+ 0.22+ 97.60+ 1.53= 100%

• About class 9, all the pixels have been correctly classified.