Coding of video sequences using hierarchical B pictures in the Joint Scalable Video Model (JSVM) for the scalability amendment of H.264/AVC has the benefit of improved rate distortion efficiency relative to other known temporal decomposition structures, besides providing temporal scalability. In the operational encoder control of the JSVM, the inter-picture dependencies within a hierarchical B picture structure are considered using a heuristic, where pictures that are more frequently used for motion compensation are coded with higher fidelity compared to pictures that are less often used for motion compensation. In this paper, we describe an approach where the dependencies introduced by motion compensation are also considered when selecting transform coefficient values. Our experimental results using an H.264/AVC conforming encoder show bit rate reductions of up to 10 % compared to the quantization method used by JSVM.