Efficient direct 2D architecture for lifted biorthogonal DWT

The paper presents a new algorithm for 2D nonseparable lifted biorthogonal wavelet transform. The algorithm is derived by factoring complementary pairs of wavelet transform filters written as (L×M) tap 2D filters. The results are efficient architectures for real time signal processing, which do not require transpose memory for 2D processing of data. The proposed architecture exploits the in-place implementation inherited from the algorithm and can take advantage of both vertical and horizontal parallelism in the direct implementation. Processing in the architecture is scheduled carefully by pipelining the lifted steps, which allows two or four times faster processing than the direct implementation. The architecture therefore allows lowering of the clock frequency by two/four. The proposed architecture operates at high speed, consumes low power and has reduced computational complexity as compared to already published filter and lifting-based biorthogonal wavelet architectures.