Because of the impacts of atmospheric turbulence, only blurred degraded image is obtained with optical detection system. A novel fast blind image deconvolution algorithm is proposed based on the specialty that most images spectral amplitude is of power law distribution, and the original estimate of the degrading process directly from degraded image can largely accelerate the speed of algorithm convergence and improve the stablity. Based on the estimation, the fine estimate of point spread function is carried on, which iteratively combined with Wiener filter image deconvolution, it ensures that the algorithm is of accurate deconvolution and universality, and finally deconvolute the image fast and effective. Experimental results show that a photoelectric detection system based on power law incremental Wiener filter (BPL- IWF) blind deconvolution of images can detect targets real time and output BPL-IWF blind deconvolution image in quasi-real time.