Near surface high resolution imaging in seismic data processing is attracting great attention. Reverse time migration (RTM) from two way wave equation is considered a good way to image the steep structure. However low frequency noise generated during the RTM imaging process is relatively high, which affects high resolution imaging of near surface . To challenge this problem, we propose in this paper a least square reverse time migration (LSRTM) method. Test results on a complex near surface velocity model indicate that LSRTM is much better than RTM for structure high resolution imaging and energy preservation.