Three-Dimensional Binary-Uncoded Video Compression Using 3DSPIHT-ZTR

In conventional video coding, motion compensation is commonly being used to remove the temporal redundancy. But the computational complexity introduced by the motion compensation can be very high, especially when the full search is adopted. The 3DSPIHT resolves this problem by replacing the motion compensation with the temporal wavelet transform, which require lower computational complexity. However, the performance of the 3DSPIHT is optimal only when Arithmetic Coding (AC) is incorporated. This contradicts the initial idea of reducing the computational complexity with the wavelet transform, because the computational complexity introduced by AC is sufficiently high. In this paper, an efficient binary-uncoded algorithm for video coding called the 3DSPIHT-ZTR is proposed. The idea is to re-introduce the use of ZTR symbol, which is initially used in the embedded zerotree wavelet (EZW) algorithm, into 3DSPIHT. In addition, the significant offspring are also encoded in a slightly different way to further enhance the performance. The performance of 3DSPIHT-ZTR is approximately 1.5 dB better than the conventional 3DSPIHT in terms of PSNR. Its performance is also comparable to 3DSPIHT with AC under certain circumstances.