Abstarct: Today, face identification/verification systems have many applications in areas such as personal identity, traffic control, and information security. These systems usually need a large databaxse of face images. Because of memory limitations and need to have high retrieval speed, we need to use an efficient compression method to compress these face images. In this task, we should preserve important face features especially in low bit rate applications. In this thesis, in an evolutionary scheme, we propose three ROI-baxsed facial image compression methods for low/very low bit rate applications. All these methods are baxsed on wavelet transform and encoding the coefficients using WBTC coding scheme. We have improved coding performance of WBTC encoder using our proposed block reordering technique which works in coefficients importance order manner. By this technique less time and bit rate is consumed to find important coefficients in encoding process. In the first method, quality of interested regions is improved by computing the residual image in wavelet domain, and encoding interested regions by Maxshift method. In the second proposed method, by using RWT transform, we compress interested and non-interested regions separately in desired bit rate. Third method is somehow a combined version of previous two proposed methods. This method uses RWT transform and residual image coding techniques and reaches to higher coding performance. Simulation results show that all proposed methods have higher compression efficiency than JPEG2000, SPIHT, and a ROI-baxsed method especially in low/very low bit rate, since by reducing bit rate, the efficiency and impact of WBTC and block-reordering techniques increase.