Two types of methods have been used to estimate motion vectors from moving image sequence : Block Matching Algorithm (BMA) and Pel-Recursive Algorithm (PRA). But, BMA suffers from limitation of estimatig complex motion since it estimates motion by matching in block by block manner. PRA overcomes this problem as it estimates recursively the motion vector for each pel. But, PRA also has another problem in esimating motion vector around the boundary of moving object since updating the motion vector is based on the gradient of intensity. In this thesis, we propose a motion estimation algorithm, Pel Matching Algorithm (PMA) which overcomes the problems of both BMA and PRA. The problem of BMA is solved by adoption the pel by pel estimating manner and the problem of PRA is solved by updating the motion vector with matching rather than gradient of intensity. We simulated our algorithm to show that PMA can estimate motion more accurately than PRA and BMA. The simulation results show that our algorithm can estimate motion more stablely and exactly than PRA and BMA.