Aiming at improving steganography security, a novel adaptive steganographic algorithm designed by non-directional filter is proposed. Firstly, the complex regions are determined by a non-directional filter in images. Secondly, a Gaussian low-pass filter is used to enhance the coherence of different pixels in those regions and then the cost function is established. Finally, according to the cost function, messages are embeded by syndrome trellis codes. Experimental results illustrate that steganographic algorithm achieves a better performance on resisting the state-of-the-art steganalysis over prior work than that of S-UNIWARD under the condition of large embedding rates and a similar performance under the condition of small embedding rates.