非极大值抑制——基于人脸检测（框出最大概率值的人脸区域）

1. 算法原理

  非极大值抑制算法（Non-maximum suppression, NMS）的本质是搜索局部极大值，抑制非极大值元素。

2. 3邻域情况下NMS的实现

  3邻域情况下的NMS即判断一维数组I[W]的元素I[i](2<=i<=W-1)是否大于其左邻元素I[i-1]和右邻元素I[i+1]，算法流程如下图所示：

  

  a. 算法流程3-5行判断当前元素是否大于其左邻与右邻元素，如符合条件，该元素即为极大值点。对于极大值点I[i]，已知I[i]>I[i+1]，故无需对i+1位置元素做进一步处理，直接跳至i+2位置，对应算法流程第12行。

    

  b. 若元素I[i]不满足算法流程第3行判断条件，将其右邻I[i+1]作为极大值候选，对应算法流程第7行。采用单调递增的方式向右查找，直至找到满足I[i]>I[i+1]的元素，若i<=W-1，该点即为极大值点，对应算法流程第10-11行。

    

3. NMS在物体检测中的应用

  物体检测中应用NMS算法的主要目的是消除多余（交叉重复）的窗口，找到最佳物体检测位置。

  

  如上图所示，人脸检测中，虽然每个窗口均检测到人脸，但仅需给出一个最有可能表征人脸的窗口。

4. 算法程序

?

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40

41

42

43

44

45

46

47

48

49

50

51

52

53

54

55

56

57

58

59

60

61

62

63

64

65

66

67

68

69

70

71

72

73

74

75

functionpickLocate = nms(boxes, overlap)

 

% Non-maximum suppression.

% In object detect algorithm, select high score detections and skip windows

% covered by a previously selected detection.

%

% input - boxes : object detect windows.

%                 xMin yMin xMax yMax score.

%         overlap : suppression threshold.

% output - pickLocate : number of local maximum score.

 

boxes = double(boxes);

 

ifisempty(boxes)

    pickLocate = [];

else

    xMin = boxes(:, 1);

    yMin = boxes(:, 2);

    xMax = boxes(:, 3);

    yMax = boxes(:, 4);

     

    s = boxes(:,end);

     

    % area of every detected windows.

    area= (xMax - xMin + 1) .* (yMax - yMin + 1);

     

    % sort detected windows based on the score.

    [vals, I] =sort(s);

     

    pickLocate = [];

    while~isempty(I)

        last= length(I);

        i= I(last);

         

        pickLocate = [pickLocate;i];

        suppress = [last];

         

        forpos = 1 : last- 1

            j= I(pos); 

             

            % covered area.

            xx1 =max(xMin(i), xMin(j));

            yy1 =max(yMin(i), yMin(j));

            xx2 =min(xMax(i), xMax(j));

            yy2 =min(yMax(i), yMax(j));

             

            w = xx2 - xx1 + 1;

            h = yy2 - yy1 + 1;

             

            if((w > 0) && (h > 0))

                % compute overlap.

                o = w * h /min(area(i),area(j));

                 

                if(o > overlap)

                    suppress = [suppress; pos];

                end

            end

             

            % xx1 = max(x1(i), x1(I(1:last-1)));

            % yy1 = max(y1(i), y1(I(1:last-1)));

            % xx2 = min(x2(i), x2(I(1:last-1)));

            % yy2 = min(y2(i), y2(I(1:last-1)));

             

            % w = max(0.0, xx2-xx1+1);

            % h = max(0.0, yy2-yy1+1);

             

            % inter = w.*h;

            % o = inter ./ (area(i) + area(I(1:last-1)) - inter);

             

            % saving the windows which o less than threshold.

            % I = I(o <= overlap);

        end

        I(suppress) = [];

    end

end