In this thesis sub-band coding of images using inter-band vector quantization is examined at rates below 1 bit/pixel. The image spectrum is decomposed into 16 uniform bands using 2-D separable quadrature mirror filters. These 16 bands are used to create 16-dimension vectores, which are coded using vector quantization (VQ). In order to reduce the computational complexity and the storage requirements, two vector quantization structures are considered. One approach is to split the vectors not only to reduce complexity but also to exploit spectral properties of the data image. In the order approach the vector quantization is done in multiple stages. Simulation results are presented at rates of 0,50 bit/pixel, 0,63 bit/pixel and 0,75 bit/pixel. A comparative analysis shows that the performance of the two systems is comparable to the basic coding structure, while the complexity is significantly reduced. Simulation results also show that split VQ with adaptive bit allocation and multi-stage VQ with tree search are not recommended at these rates. The best trade-off between complexity and performance is achieved with the inter-band multi-stage VQ scheme using the sequential search procedure. At last, the use of inter- band VQ is examined when the dominant sub-band is separately coded using intra-band VQ. When the image spectrum is decomposed into 16 bands, this scheme has a performance comparable to the basic coding structure while the complexity and storage reduction is equivalent to the inter-band multi-stage VQ technique.