A data caching method for a video decoder includes: receiving a data request from the video decoder, wherein the data request is for requesting one or more reference pixel blocks to decode a present pixel block, and the data request contains one or more reference motion vectors each corresponding to a reference pixel block; determining, for each reference motion vector in the data request, whether the reference pixel block corresponding to the reference motion vector has been cached in a cache memory which caches a plurality of reference pixel blocks; and if the reference pixel block corresponding to the reference motion vector has not been cached in the cache memory: A) determining an expected usage number and a historical usage condition for each reference pixel block which has been cached in the cache memory, wherein the expected usage number indicates how many times the cached reference pixel block will be used in decoding of a predetermined number of subsequent pixel blocks after the present pixel block, and the historical usage condition indicates a least recently used record of the cached reference pixel block; and B) selecting and deleting one of the cached reference pixel blocks according to the expected usage numbers and historical usage conditions of the cached reference pixel blocks, caching a reference pixel block corresponding to the reference motion vector from an external memory into the cache memory, and providing the reference pixel block to the video decoder.