The proliferation of wireless technology, mobile computing, and increasingly sophisticated video codecs have fueled the sharp increase in demand for user-centric applications involving wireless video streaming. Consequently, it is becoming progressively more important to ensure that wirelessly transmitted video signals are delivered with optimal perceptual quality to the end-user. This is in contrast to traditional QoS, which aims to optimize system and packet-level performance, but not necessarily perceptual quality. In order to drive towards this goal, one of the most important evolutions to modern video streaming systems will be the adaptive allocation of both wireless and video coding resources at the transmitter - ideally according to a cost function designed to optimize some measure of the received perceptual video quality. Unfortunately, it is not at all likely that a transmitting station will have access to the video signal received by the end-user, and therefore, this precludes the possibility of using such a perceptually-oriented cost function. Indeed, current resource allocation strategies are based on simplistic packet-statistic-based estimates of the received video quality, and one can easily demonstrate that this results in suboptimal perceptual video quality to the end-user. In this paper, we propose a simple, but powerful technique which allows a transmitting station to obtain an accurate estimate of the video signal seen by the recipient using cross-layer feedback between the wireless MAC and the video encoder, thus enabling the transmitter to use more perceptually oriented resource allocation strategies. We support the proposed technique through both simulation and experiments involving a database of over 420 distorted H.264 AVC compressed video sequences.