Abstarct: Digital medical images like X-Ray are widely used in diagnosis. One of the considerable advantages of digital medical images for patients is ease of storing and transmission of them. Obviously, less volume of images cause to faster compression and transmission. Furthermore, they need less space for storing. The most significant issue is attempting to preserve quality of important features of tooth adequately, especially in low bit-rates. In this thesis two ROI-baxsed methods are proposed in a progressive manner to compress dental images in low bit-rates. Wavelet transform and SPIHT encoding are the basis of both presented methods. In both methods, at first, region of interest should be specified, to do this we investigated three methods and finally chose and modified one of them. These three methods are: wave propagation method, active contour method and joining up boundaries method. Its performance in finding filled regions has been improved by applying median filter to eliminate noise and perform some changes in tracing method on detected edges by canny operator, so that the proposed method is able to join up boundaries and determine regions of interest with adequate accuracy. Finally third method was chose, and used to compress. In first method, region of interest and background were encoding separately but in maxshift method they were processed together. In maxshift method compression rate is higher but least squares error in ROI in this method is higher than first method. Processing speed in SPIHT method is higher, because in encoding step of maxshift, the bit stream is generated several times so it cause to low speed in maxshift method.