Frequency domain down-conversion of HDTV using an optimal motion compensation scheme
✍ Scribed by Anthony Vetro; Huifang Sun
- Publisher
- John Wiley and Sons
- Year
- 1998
- Tongue
- English
- Weight
- 231 KB
- Volume
- 9
- Category
- Article
- ISSN
- 0899-9457
No coin nor oath required. For personal study only.
✦ Synopsis
In the MPEG-2 Test Model 5, the down-conversion of conversion. To reduce the cost of this system, some researchers an interlaced sequence is obtained by prefiltering and subsampling have considered performing the down-conversion within the deeach field of the image sequence after it has been fully decoded.
coding loop [4][5][6][7]. Although the quality of the picture will be Although the quality is very good, the cost of such a system is quite compromised, significant reductions in the amount of memory high owing to large memory requirements. As a result, low-resolution can be realized. This type of system will be referred to as a lowdecoders have been proposed to reduce some of the costs incurred resolution decoder (LRD) and is shown in Figure . by this scheme. Here, incoming Discrete Cosine Transform (DCT)
The two processes in the LRD which affect the quality of the blocks are subject to a down-conversion process within the decoding down-converted image sequence are the actual down-conversion loop; hence, the motion compensation is performed using the downand the motion compensation (MC). One of the main objectives converted images. In past work, it has been proven that the optimal filters for performing this motion compensation are intimately related of these processes is to keep prediction drift to a minimum, where to the method of down-conversion. Therefore, the choice of downprediction drift is defined as the successive blurring of forward conversion filter is viewed as the primary variable affecting the quality predicted frames within a group of pictures (GOP). This blurring of the down-converted sequence when such an optimal motion comeffect is mainly owing to nonideal interpolation of sub-pel intensipensation scheme is considered. In the conventional method of freties and also the loss of high-frequency data within a block. Not quency domain down-conversion, the 4 1 4 low-frequency coeffionly is prediction drift a disturbing artifact within the GOP, but cients are extracted from each 8 1 8 block. Two problems arise from over a series of GOPs, a periodic pulsing can be observed owing this method: First, the discarding of high frequency data will introduce to the increase in quality from the last P-frame to the next Ia disturbing amount of drift; and second, severe blocking artifacts frame. This is a very serious artifact, and if not treated properly, will result in areas of large motion. To remedy the drift problem, a the quality of the decoded sequence can not be deemed acceptnew method of down-conversion which better preserves high-frequency data is presented. This method is referred to as frame-based able. In [8], expressions for the optimal set of filters to perform frequency synthesis. Then, to overcome the blocking artifacts, the the required low-resolution motion compensation have been deprevious method is extended to a field-based frequency synthesis rived. The optimal least-squares solution effectively minimizes method. Our simulation results clearly indicate that the amount of the drift for any method of down-conversion which operates on drift can be significantly reduced by retaining high-frequency data a single macroblock. The results show that the optimal set of and that severe blocking artifacts are eliminated by using the fieldfilters are intimately related to the method of down-conversion.
based method. In addition, the quality achieved by the proposed field-