In this paper, a new 3D intermediate views reconstruction technique using an adaptive disparity estimation algorithm is proposed and its performance is analyzed by comparison to that of the conventional disparity estimation algorithms. In the proposed algorithm, in order to effectively synthesize the intermediate views the matching window size is selected according to the feature value of the input stereo image. By doing this, the mismatching probability of the disparity can be reduced through coarse matching in the similar area and fine matching in the area having large feature values such as in the edge part of object. From some experimental results, it is found that the proposed algorithm improves the PSNR of the reconstructed intermediate views about 3-4 dB on the average than that of the conventional algorithms.