Lossless image compression can preserve all information of the original image and is useful for security and medical image processing. In this paper, we apply improved adaptive arithmetic coding to further improve the coding efficiency of context-based adaptive lossless image coding (CALIC), which is a well-known lossless image compression algorithm. We apply the techniques of mutual learning, initialization for the frequency table, and the increasingly adjusting step to improve adaptive arithmetic coding and apply them in the CALIC algorithm. Simulations show that, with the proposed coding techniques, the performance of CALIC for lossless image compression can be obviously improved.