A motion vector estimation apparatus that can reduce an amount of calculation for video coding processing and thus achieve a higher bit rate and lower consumption power, while contributing to improvement in image quality and coding efficiency include: a reduced picture generation unit which generates a reduced current picture to be coded and a reduced reference picture; a region partition unit which partitions a reduced current picture to be coded into regions and generates reduced region images; a region motion vector estimation unit which estimates a region motion vector of a reduced region image; a confidence level calculation unit which calculates a confidence level of a region motion vector; and a block size narrowing-down unit which narrows down candidate block sizes so as to determine a block size to be used for coding a current block to be coded, based on a region motion vector and a confidence level of the region motion vector.