A Parallel Method for Stego Image Feature Extraction on Multicore CPU

At present, the key techniques of the universal steganography detection include image feature extraction and classifier construction. With the structure of features for steganography detection being more and more complex, the computation of image feature extraction algorithms constantly increases, which becomes the most time-consuming part of image steganography detection. In this paper, we focus on the parallelization method for stego image feature extraction on multicore CPU system. By overcoming some disadvantages of the original OpenMP parallel method, we propose a feature extraction method that uses thread-level task parallelism, which firstly constructs a lock-free task queue for task threads, secondly reduces thread synchronization overhead and finally solves false sharing issue and sets thread affinity scheduling to improve performance. Results of the experiment show that the proposed parallel method works out good speedup performance on the dual-core and quad-core systems. Compared with the original OpenMP method, our method gains better speedup that is 1.2% and 3% faster respectively, and that improves the practicality of the universal steganography detection.