This paper presents a massively parallel hybrid, bispectrum speckle imaging (SI) and multi-frame blind deconvolution (B-MFBD) image reconstruction technique for anisoplanatic, long horizontal path imaging. We use a fast initial estimate, generated using parallel bispectrum SI technique, to initialize the parallel multi-frame blind deconvolution (MFBD) algorithm. The proposed hybrid B-MFBD technique, through massive parallelization, significantly speedups both existing iterative bispectrum SI and MFBD parts of the hybrid algorithm. We report a speedup factor of 7.31 with respect to sequential implementation of the same algorithm for an image size of 256 × 256.