An Illumination Adaptive Video Coding Scheme for In-Vehicle Video Applications

With the advance of intelligent vehicle systems, drivers or passengers can keep interaction with people in fixed offices or other vehicles through visual communications. However, the illumination variations due to the changes of environments or weather conditions may significantly change the appearance of in- vehicle videos. Accordingly, the compression efficiency is much reduced even though the bandwidth of such wireless communications has been quite limited. There is pretty few previous work designed for efficient in-vehicle video compressions. Thus, we propose an illumination adaptive video coding scheme for in-vehicle video applications. Since human faces are usually the most visually attended regions in such applications, this scheme consists of illumination correction, face detection, and the visual attention based video codec. The proposed illumination correction strategy combines the advantages of the single-scale Retinex (SSR) and the weighted histogram separation (WHS). The experimental results show that our illumination correction strategy effectively improves the face detection performance of in-vehicle videos. Moreover, the subjective visual quality of the proposed scheme outperforms that of H.264 with rate control since our scheme allocates bits by incorporating the psychovisual characteristics.