提出一种以线为单位进行帧内预测编码的新方法,从预测及变换两方面提高帧内编码的性能。首先在宏块内部以水平或垂直方向的线为单位,分别参考相邻的行或列像素进行多个方向的帧内预测,减少了预测像素与参考像素间的距离,从而提高预测精度;其次,在原有的4×4离散余弦变换(DCT)基础上补充了1×16的一维DCT,并以宏块为基本单位灵活选择变换方式。实验结果表明,在同样的重建图像峰值信噪比(PSNR)下,本文方法与H.264/AVC相比,平均码率节省约为6.7%。
A novel line-based intra coding method is presented to improve the coding efficiency by both prediction and transform. Firstly, the spatial multi-prediction is performed line by line in horizontal or vertical direction within a macroblock using the pixels of the adjacent column or row as the reference, which improves the prediction accuracy by shortening the distance between the reference and current pixels. Secondly,on the basis of the original 4 X 4 discrete cosine transform (I)CT),a 1 X 16 one-dimen- sional IRET is introduced as a candidate,and the transform is selected adaptively at the level of macrob- lock. Experimental results show that compared with the H. 264/AVC, the proposed method reduces the bit rate by 6.7% approximately at the same peak signal to noise ratio (PSNR).