Adaptive Transformation for Robust Privacy Protection in Video Surveillance

Algorithm 1

Adaptive data transformation.

Input:  Original Video: 𝑉={𝑓₁,𝑓₂…}

  and set of static evidence regions: 𝑅^𝑠={𝑟^𝑠₁,𝑟^𝑠₂…};

Output:  Transformed Video: 𝑉′={𝑓_1′,𝑓_2′…}

Description:

   1:  for all 𝑓_𝑖∈𝑉  do

   2: // Detect dynamic evidence regions

   3: 𝑅^𝑑=𝐷𝑆𝑅(𝑓_𝑖);

   4: if 𝑅^𝑑==𝐸𝑀𝑃𝑇𝑌  then

   5:  if 𝑅^𝑑==𝐸𝑀𝑃𝑇𝑌 for previous 𝜔 frames then

   6:   𝑓^𝑡_𝑖=𝐶𝑂𝑃𝑌(𝑓_𝑖);

   7:   for all 𝑟^𝑠_𝑗∈𝑅^𝑠  do

   8:    𝑓^𝑡_𝑖=S𝑇(𝑓^𝑡_𝑖,𝑟^𝑠_𝑗);

   9:   end for

10:  else

11:    //Do the global transformation

12:   𝑓^𝑡_𝑖=𝐺𝑇(𝑓_𝑖);

13:  end if

14: else

15:   //Foreground detected

16:   //Transform static regions

17:  𝑓^𝑡_𝑖=𝐶𝑂𝑃𝑌(𝑓_𝑖)

18:  for all 𝑟^𝑠_𝑗∈𝑅^𝑠  do

19:   𝑓^𝑡_𝑖=𝑆𝑇(𝑓^𝑡_𝑖,𝑟^𝑠_𝑗);

20:   𝑀𝐴𝑅𝐾(𝑓^𝑡_𝑖,𝑟^𝑠_𝑗)=𝑇𝑅𝑈𝐸;

21:  end for

22:  //Transform dynamic regions

23:  for all frame 𝑟^𝑑_𝑘∈𝑅^𝑑 do

24:    //Calculate parameters for space variant operation

25:   𝑃𝑅𝑀=𝑃𝐸(𝑟^𝑑_𝑘)

26:   if !𝑀𝐴𝑅𝐾(𝑓^𝑡_𝑖,𝑟^𝑑_𝑘) then

27:    𝑓^𝑡_𝑖=𝐷𝑇(𝑓^𝑡_𝑖,𝑟^𝑑_𝑘,𝑃𝑅𝑀);

28:   end if

29:  end for

30:   //Data transformation over

31: end if

32:  //Copy frame to output frame sequence

33:  𝑓_𝑖′=𝐶𝑂𝑃𝑌(𝑓^𝑡_𝑖);

34:  end for

35:  return Transformed frame sequence 𝑓_𝑖′ as Video 𝑉′;