A context adaptive tri-element Run length coding algorithm and an Exp-Golomb coding alogorithm were introduced.The Run-length coding was used to model the image wavelet coefficients and the Region of Interest(ROI) mask and to yield tri-element codes.Then, the conventional Exp-Golomb coding was expanded to encode tri-element codes from Run-length coding and to carry the ROI mask together. Based on the two algorithms above,a high efficient algorithm to distinguish the ROI and background was obtained and a ROI codec framework was proposed. The framework includes 5/3 wavelet transform, wavelet domain mask generation, non-uniform optimal quantization, context adaptive tri-element Run-length coding and generalized Exp-Golomb coding. The tri-element Run-length coding in this algorithm is brief, and the expanded Exp-Golomb coding can be expressed by a closed formula. Experimental results show that the algorithm supports multiple arbitrary-shaped ROI and the adjustability of the ROI is prior to that of the background region. Furthermore, it achieves higher compressing performance as compared with the BbB-shift based SPIHT compression algorithm.