Embodiments of the invention provide a video communication system, device and method on the basis of feedback reference frame. The method includes: receiving, by a video encoding end, video frame information fed back by a video decoding end, determining a reference frame for encoding current video image according to the video frame information; encoding the current video image according to determined reference frame, and sending a video frame formed after the encoding; decoding, by the video decoding end, received video frame, and feeding back the video frame information to the video encoding end through a network and according to decoding result. By adopting embodiments of the invention, network data flow may be reduced. Meanwhile, phenomenon which may cause greater impact on the network, when it is necessary to send larger data amount since frame-loss or failed-decoding, may be avoided.