Most video coding schemes are based on the independent processing of nonoverlapping image blocks. An important disadvantage of this approach is that blocking artifacts may be visible in the decoded frames, particularly for very low bitrate coding. The LOT(lapped orthogonal transform) is one of the effective techniques in reducing the blocking artifacts.
We investigate the performances of LOT for still image coding and the CLT(complex lapped transform) for moving image coding. For still image coding, it is shown that LOT/CVQ(classified vector quantization) is effective for progressive image transmission, taking advantage of reduced blocking effects. For moving image coding, the characteristics of motion estimation by using CLT(complex lapped transform) is investigated. The motion estimation is based on cross-correlation function that can be calculated in the frequency domain after CLT for overlapped regions. It is shown that the motion vector field obtained by CLT is more homogeneous than others.
For application of the CLT-based motion estimation to the standard low bitrate video coding, e.g. H.263, a new decision method based on the same cross-correlation function is proposed for the Coding/No-Coding decision for each macroblock. It is found more robust and effective than the existing decision method that is applied after transform and quantization. By simulation to several test image sequences, it is shown that the proposed motion estimation and coding block decision by CLT transform give better performance than the conventional method in the standard low bitrate video coding scheme.
