Current link adaptation algorithms for IEEE 802.11 WLANs exhibit behavior which makes them unsuitable for transmission of multiple simultaneous real-time uplink video streams. First, these algorithms do not consider the properties of the video codec, and hence, are unaware of the impact of PHY rate selection on the perceptual quality of the received video. Second, they do not differentiate between channel errors and collisions, and hence severely malfunction when the collision probability is non-negligible. In this paper, we propose a link adaptation strategy that not only optimizes the perceptual quality of the received video, but also maintains network stability by preventing catastrophic failure due to collisions. We show that switching to a lower PHY rate improves the SNR/BER performance, but increases channel contention (and hence the collision probability). Then, we use this information plus knowledge of the video codec and network transport protocol to estimate the received perceptual video quality at the current and adjacent PHY rates. The PHY rate that yields the best perceptual quality is chosen for each Group of Pictures (GOP). We support the proposed algorithm through experiments with real wireless cameras on which we have implemented our algorithm.