Overtaking maneuver on roads without a clear view can cause serious accident. Hence, real-time video streaming deliveries between two vehicles over Vehicular Ad Hoc Networks (VANETs) have been proposed as an overtaking maneuver assistance system. The system is also known as a V2V see-through system since it sends real-time video scenes seeing by a front vehicle to the driver of the back vehicle to help him decide on overtaking maneuver. However, sending real-time video streaming over VANET is still problematic due to delay and packet loss. Several solutions have previously proposed to fix the problems, but all of them still have some drawbacks. So, we propose in this paper a cross-layer optimization of video streams (video frame skipping, video transcoding and frame rate reduction) for overtaking environments. The proposed approach is based on IEEE 802.11 EDCA. Experimental results on our approach using OMNET++ have demonstrated some favorable points, such as bandwidth waste avoidance and acceptable video latency.