Architectures for the DCT
Due to the importance of the two-dimensional (2-D) DCT in digital image processing, particularly in video compression, various algorithms and architectures have been proposed for its implementation. These methods can be divided into two main categories. The first category includes techniques from linear matrix analysis and decomposition. Lee’s fast DCT is a typical example of this category. The second category includes algorithms from polynomial and number theory. For example, Winograd’s algorithm fits into this class. Both categories can be further partitioned into algorithms that either use or do not use row-column decompositions. For example, with a row-column decomposition, the 2-D DCT can be computed using one-dimensional DCTs, allowing for simpler implementations. Algorithms that do not use row-column decompositions are more computationally efficient but require complex hardware.
KeywordsVideo Compression Video Codec Parallel Multiplier Distribute Arithmetic Pixel Rate
Unable to display preview. Download preview PDF.