A video encoding apparatus comprises a motion vector detector to detect a motion vector of an input picture referring to a reference picture, and a predictive encoder to perform forward predictive encoding and bidirectional predictive encoding using the motion vector and the reference picture, the forward predictive encoder subjecting the macroblock of the forward predictive encoded picture to a variable length encoding in not_coded mode when a correlation between the macroblocks of the forward and bidirectional predictive encoded pictures and the reference picture which are located at the same position is high.