Video encoding device using predicted motion vector for blocks of different size

A method implemented by a computing device having a pipelined processor performing operations comprising:

• determining a search start point of a motion vector for each block from a prediction vector, the prediction vector being a motion vector determined with a plurality of motion vectors for a plurality of surrounding blocks adjoining each block and in accordance with a preset rule,
• searching for the motion vector for an encoding target region from the search start point;
• calculating the prediction vector by temporarily determining a division pattern for the encoding target region from a division pattern for an encoded macroblock in a target image other than an encoding target macroblock when a division pattern for the encoding target macroblock having each block undefined in calculating the prediction vector in accordance with the preset rule;
• calculating the prediction vector by substituting a motion vector for a block in the target image other than the encoding target macroblock when at least one of the plurality of motion vectors for the plurality of surrounding blocks cannot be acquired;
• encoding a difference between the prediction vector and the motion vector for the encoding target region that has been searched for,
• wherein a temporary determination includes a determination whether to integrate each pair of blocks that has an adjacent relationship when the target macroblock is divided in accordance with a predetermined division pattern, a division pattern that is same as or most similar to a division pattern resulting from a combination of a plurality of determinations is determined from a plurality of acceptable division patterns, and each of the plurality of determinations is made such that each pair of blocks have a same integration state as a pair of blocks out of a plurality of blocks into which a surround macroblock whose division pattern has been determined is divided in accordance with the predetermined division pattern, the pair of blocks including the plurality of blocks arrayed in a same direction as each pair of blocks and adjoining the each pair of blocks;
• and
• determining a substitute motion vector, from a plurality of candidates, on a basis of a result of a comparison between a threshold and a variance value of a plurality of respective motion vectors for at least three blocks included in a plurality of encoded macroblocks in the target image other than the encoding target macroblock when at least one of the plurality of motion vectors for the plurality of surrounding blocks cannot be acquired.