Abstract: This study proposes a novel segmentation based tree structured algorithm for efficient compression of digital images. The presented research combines the recent segmentation based coding approach, namely the binary space partition scheme and the popular Geometric Wavelet Coding Method to capture the curve singularities in a more effective way and to provide the sparse representation of the image. For partitioning the image domain in the BSP scheme, polar co-ordinate representation of straight line is used which improved the choice of bisecting lines available for partitioning thereby enhancing the probability of reducing the error functional. A rate-distortion optimization process is performed prior to encoding where a New Pruning algorithm is tried to prune the BSP tree and achieve the desired bit rate. The relative practical competency of this hybrid technique is investigated and the results are compared with state-of-the-art wavelet coders, recent segmentation based algorithms as well as the Original Geometric Wavelet Coding algorithm. This technique provides outstanding results in terms of rate-distortion compression by taking advantage of the curve singularities in the image. The algorithm is applied individually on tiled regions of the image rather than on the entire image. The results report a gain of 2.46 dB over the EZW algorithm and 1.74 dB over the SPIHT algorithm at the bit-rate 0.03125 bpp. Researchers also show that the presented algorithm reports a gain of 1.2 dB over the Original GW Method at the compression ratio of 256 for the Lena test image. The technique provides remarkable results in terms of rate-distortion compression by taking advantage of the edge singularities in the image. The improved GW algorithm was simulated using the 2010 version of MATLAB on still images of Lena and cameraman to validate its performance. The algorithm is highly complex in computation and requires enormous time for execution.