A ship target detection scheme based on optical remote sensing images of contrast box filtering over the standard deviation feature planes is presented. Firstly, local intensity variance is selected as the detection feature to characterize ships of different intensities uniformly and eliminate the influence of varying sea average intensity at the same time. Secondly, on the two-dimensional detection feature plane, local detection thresholds are estimated self-adaptively by contrast box filtering and the spatial structure information of ship targets is combined to identify the locations of ship candidates in the intensity image. Thirdly, verification of each ship candidate is performed in combination with the prior knowledge ship feature model to remove the false alarms, and then the final results are outputted. Experiments conducted on the optical images demonstrated that the ship target detection probability can be as high as 99.5% and the false alarm rate is kept at 5%.