Correlation-based normalized partial distortion search algorithm for motion estimation

A modified partial distortion search algorithm considering the neighboring-block correlation property is proposed for fast motion estimation. The motion vector information of neighboring coded blocks is used to predict the possible occurrence region of the global-minimum-distortion position of the current block. In addition, a dynamic searchrange decision algorithm Is also proposed for automatically changing the size of the search range. Afterwards, the normalized partial distortion search Is performed in the selected region instead of the whole search window. Through the proposed algorithms, the computational complexity can be significantly reduced with slight objective quality degradation.