Multimedia data transmission over sensor networks is crucial in monitoring and surveillance applications, such as object tracking. In these applications, several sensor nodes are equipped with cameras, and placed on pre-selected locations. When an event occurs, cameras around event capture images and transfer them to the sink over a multi hop path. Since sensor nodes operate with limited CPU, memory and bandwidth, resource utilization is needed for feasible processing and transmission of useful parts in an image. In this study, we provide network performance trade-offs for an object tracking application. In our experimental setup, the application utilizes costeffective coding of image macro-blocks into "high" and "low" classes based on the object information they contain. We then investigate resource trade-offs for achieving the highest object transmission rate when only "high" blocks are injected into the service differentiated network. We conclude with some remarks on the tuning of network buffer length and retransmission count parameters, for the optimization of macro-block reliability in the given object tracking scenario.