Signal compression through spatial frequency-based motion estimation
Integration, the VLSI Journal
| Hi-index | 0.43 |
One of the most efficient image compression techniques investigated is the discrete cosine transform (DCT). The authors present a method where, through the use of pruning, the speed of implementation of the algorithm is dramatically increased. They describe its use in image compression. The advantage of the pruning algorithm in such cases being that, only a small proportion of the DCT values are retained hence why calculate the values which are not retained. Results show that for an 8 by 8 image block it is only necessary to calculate a 4 by 4 subset of the DCT values to retain an acceptable image quality, this gives a considerable speedup in calculation time compared with calculating the full 8 by 8 DCT block. Another advantage of the technique is that parallelisation can be achieved using the data partitioning method. For the JPEG standard, the image is split into 8 by 8 blocks and the DCT of each block calculated, each block is independent of its neighbours and hence can be calculated in parallel, as with all parallel applications data transfer times are crucial, here the pruning algorithm has a second advantage in that only a subset of the results of the DCT need be transferred back to the root process. Results are presented for its application to the JPEG standard. They discuss both the effects of pruning on the image quality and the effects of pruning on parallelisation and speedup