Objective To sovle the defects such as poor imperceptible performance, low randomness of the confidential information embedding path, difficulty in fighting against blind steganography attacks, and small capacity of infimation hiding in current image steganography mechanism. Methods The image steganography based on pixel optimization and graph random embedding path was proposed in this paper. Carrier image was converted to frequency domain integer coefficients based on HIWT mechanism, and lossless steganography was achieved. The hiden information length estimation model was constructed by defining the image histogram modifying rule to find the frequency domain representation of image blocks. Then the pixel optimization adjustment mechanism was designed by introducing the LSB substitution to reduce the error between carrier image and stego-image for increasing steganography capacity. Random embedding path of frequency domain integer coefficients were designed by embedding graph to improve the imperceptible performance. Finally, the extraction mechanism was designed to obtain the stego-image. Results The simulation results on two large image databases showed that in comparison with current image steganography mechansim, the imperceptible performance and anti-detection of the proposed method were stronger, which effectively reduce the ladder effects; and the high capacity of this mechanism was higher. Conclusion The algorithm proposed in this paper could hide the images in a safe and lossless way, and can effectively protect the reliable transportation of images.