Conventional energy detection (CED) is the optimal signal detection in uncorrelated noise field. Subband peak energy detection (SPED) further improves the display resolution of CED by exploiting the spatial coherence of the energy peaks. However, under low signal to noise ratio (SNR) condition or where there are few snapshots, the performance of weak target detection submerged in strong interferences by SPED is still unsatisfying. An improved SPED method is proposed in this paper. The improved SPED algorithm exploits the spatial continuity of the signal besides the spatial coherence. Restricted by the resolution of beamforming method, the local peaks of the target would not occur in the exact same azimuth for each frequency bin. By integrating these scattered peaks, the energy of the weak target will be enhanced. A double detection is then implemented to pick out the noise peaks. Experiments are conducted with simulated data to validate the effectiveness of the proposed method. Results show a better display resolution of the proposed method with a better detection performance of the relatively weak target, especially under poor conditions when the number of snapshots is small or SNR is low.