Automatic Representation and Segmentation of Video Sequences via a Novel Framework Based on the D-EVM and Kohonen Networks

Algorithm 9

Algorithm for the convolution process to generate set DCValues.

   Input: The D-EVM for the mask, temporal position of the final frame, and the width and height of the frame.

   Output: The DCValues set that contains the values of DC descriptor for every possible sub-animation.

(1)  Procedure  AnimConv  (D-EVM mask, integer endFrame, integer width, integer length)

(2)     EVM currentResult; // Current intersection result

(3)     EVM frameSeq;

(4)     integer , maxTimeShift;

(5)     integer , maxXShift;

(6)     integer , maxYShift;

(7)     real maskDC;

(8)     Set of reals DCValues;

(9)     integer ; // Counter for every time shift.

      / Obtaining the frame sequence required for the intersection operation. /

(10)      frameSeq ←  maskAnimSections (mask, endFrame);

      / Obtaining the maximum shift for the time, and dimensions. /

(11)      maxTimeShift ← endFrame − mask.timeLength + 1;

(12)      maxXShift ← width − mask.xLength + 1;

(13)      maxYShift ← length − mask.yLength + 1;

(14)       ;

(15)       while  do

(16)          while y ≤ maxYShift do

(17)             while x ≤ maxXShift do

(18)               currentResult ← maskIntersection (mask, frameSeq);

(19)                 ← discreteCompactness (currentResult, mask.LcMin, mask.LcMax);

(20)               DCValues.addDCValue ();

(21)               mask.EVMTraslation (1, 1);

(22)                ;

(23)                ;

(24)            mask.dimReset (1);

(25)            mask.EVMTraslation (2, 1);

(26)             ;

(27)             ;

(28)        mask.dimReset (2);

(29)        mask.EVMTraslation ();

(30)        frameSeq ← maskAnimSections (mask, endFrame);

(31)        ;

(32)        ;

(33)        ;

(34)       return DCValues