Motion vector correction based on the pattern-like image analysis

Abstract

We propose a motion vector correction algorithm to correct the erroneous motion vectors due to the failures of conventional motion estimation (ME) based on block matching algorithms (BMA). The proposed method is based on the pattern-like image analysis that corrects the erroneous motion vectors by considering multiple possible motion vectors and by excluding the wrong motion vectors. Also, it does not significantly increase computational complexity since it uses the estimated motion vectors and the sum of absolute difference (SAD) previously obtained from the conventional ME. Experimental results indicate good performance of the proposed scheme combined with frame rate up-conversion system.