An image coding scheme with activity-based macro block classification and half-pixel precision motion estimation

We propose an image coding scheme with the following two features: (1) all macro blocks in one picture are classified into two types according to their activity, and are coded in different ways, thus, making the encoding process more effective with only a little decrease in the quality of the reconstructed images. (2) With half-pixel precision motion estimation, the quality of the reconstructed images is improved, at the expense of a little increase in the computation complexity. The discrete cosine transform (DCT) is a kind of orthogonal transform, which has been widely used in image coding because it is easy to perform and is irrelevant to input signal. The quantized 2-dimensional (2D) signal must be transferred into a 1-dimensional (lD) signal before being encoded with a variable length coding (VLC). A `0´ run-level coding, which is in fact a kind of 2D Huffman coding, is used