In this paper, lossless image coding techniques are discussed first, then a lossless image coding method based on block direction prediction and context is presented. It removed the redundancy of the image in spatial domain by block direction prediction and context-based error modeling. In the proposed method, an image is first partitioned into blocks. Then, a direction predictor that results in minimum prediction error is adaptively selected for each block. After that, context selection and error feedback are applied to further reduce the entropy. A fast and efficient Rice coder is applied finally for the residual image coding. The experimental results show that the proposed method significantly outperforms over JPEG (joint of picture expert group) lossless mode and FELICS (fast and efficient lossless image compression), and it obtains comparison ratio superior to CB9, LOCO-I, even UCM.