Feature point based motion vectors detecting apparatus

The present invention by using a motion estimation method based on the feature point to an apparatus for detecting a motion vector, the motion vector detection device for detecting the initial motion vectors for all the feature points; By selecting each as a subject feature point in sequence of the feature points, the feature point register to form the four feature points so surrounded by the lozenge shaped previous frame feature point that is the subject of the respective neighbor; And a motion vector storage register for storing the initial feature point in the first set; A motion vector displacement generator by giving the N number of displacements to the initial motion vectors for the subject feature point of each generate N number of candidate motion vectors for each of the subject feature point and; The initial motion vector and Sy around by said four feature points of the four feature points by using the N candidate motion vector of the subject feature point and detect a motion vector of the second set of pixels contained in a lozenge previous frame, a predicted water chestnut model the pixel within the current frame generated by using the motion vector of the second motion compensator set to assign his pixel value; Rhombic present frame and therefore corresponds to calculating the difference between the predicted current frame, each lozenge for generating the sum of difference for each of N previous frames each lozenge subtracter and an adder, and; A selector for selecting one of a motion vector corresponding to the sum of a car carrying a minimum difference from the N candidate motion vectors.