The work aims to design the fast image encryption algorithm based on weighted histogram bit confusion and fractional order chaotic diffusion, for the purpose of solving theses defects as low anti-plaint attack induced by ignoring the characteristics of plain pixels during the scrambling and diffusion, and low security caused by using the same encryption function during the whole pixel diffusion in current image encryption technology. The algorithm made full use of the pixel value of the entire plaintext, and embedded it into the whole scrambling and diffusion phase, and different encryption functions were used to diffuse different pixels in the diffusion process. Firstly, combined with Logistic and Tent map, a new low-dimension chaotic system was constructed by means of the nonlinear combination thought and its chaotic performances were analyzed. The pixel weighted histogram was established by taking into account the pixel value of the input plain, and the initial value of the composite chaotic system was generated by means of the external key. The random sequence was outputted by iteration. Then, each pixel of the plaintext was extended on the level of the plaintext, and the plaintext was confused by the discrete random sequence. Subsequently, the fractional order Logistic chaotic map was constructed by embedding fractional order theory into Logistic map, and the chaotic array was outputted to iterate it with the weighted histogram of the pixel. The different diffusion model was designed by classifying the pixels of the confusion cipher and combining the chaotic array to differently encrypt three kinds of pixels. The test results showed that the proposed encryption scheme had a stronger ability to resist plaintext attacks, and the pixels of output cipher was more uniform compared with the current chaotic encryption algorithm. With higher security and efficiency, the proposed encryption technology can better protect the images during their safe transmission in the network.