Recently, a study of colorization algorithm for monochrome still image has prospered. Considering actual applications, an extension to monochrome motion picture becomes an important subject. In our previous work, a colorization method for monochrome motion picture was proposed. The method consists of two steps. The first step detects displacement vectors by a block-matching of optical flow scheme between time-sequential adjacent frames. The second step transfers color between corresponded pixels. In the first step, error detection of displacement vectors often occurs in the conventional method. The error accumulates from a previous frame to the next frame, and it causes a fall of colorization accuracy. This paper proposes an improved colorization algorithm for motion pictures. In the proposed algorithm, only reliable displacement vectors, which can be evaluated by the luminance difference between corresponding blocks, are used for color transfers between adjacent frames in the second step. In the third step, the reliable transferred color will be propagated to remaining pixels by using a colorization technique for still images. Since the proposed algorithm ignores displacement vectors with low reliability, it becomes possible to decrease error colorization. This paper discusses the performances of the proposed colorization algorithm with experimental results for natural motion pictures.