In traditional video coding standards, such as MPEG-x and H.26x, MSE (Mean-Square Error) is in general the standard search metric for motion estimation. Wyner-Ziv video coding is a new coding paradigm different from the traditional video coding standards by shifting from the encoder to the decoder the process of predicting the current frame based on motion estimation and motion compensation. However, in Wyner-Ziv video coding almost existing motion compensated temporal interpolation (MCTI) algorithms still use the MSE as the search metric. Does the MSE really fit for MCTI, which is carried out at the decoder of the Wyner-Ziv video coding? This paper tries to answer the question and a new search metric combining the structure similarity with MSE is proposed to improve MCTI performance.