2D Lifting Based Discrete Wavelet Transform Architecture Using Modified Reduced Hardware Algorithm

摘要

Lifting Based Discrete Wavelet Transform (LDWT) is the proficient mode to generate biorthogonal wavelet transforms and as well to work out classical wavelet transforms. This Modified Reduced Hardware Algorithm (MRHA) is based on a flipping technique in order to reduce arithmetic components with a straightforward control path. To achieve one multiplier delay of critical path and 100 percent hardware utilization efficiency the lifting data flow is performed using parallel computations. The proposed MRH algorithm is varied in accordance with the mathematical modifications.MRH algorithm uses associative principle which is applied to the mathematical modifications of the original lifting steps to reduce the number of arithmetic components required to perform compression. This architecture follows the basic lifting steps like splitting, predicting and updating but merges the predictor with the updater. By performing parallel operation using the two-input, two-output architecture the high pass and low pass values can be calculated. DWT provides high coding efficiency and better quality of image restoration when assimilate with the conventional Discrete Cosine Transform.As a result, it is a speed competent, least hardware regular design with minimum memory requirement. Because of these competences it can be useful to several real-time visual setups viz. JPEG2000, MPEG-4 still texture object decoding, etc. The MRHA based architecture is verified in Lab VIEW tool that shows that the architecture splits the image into four sub-bands namely LL, LH, HL and HH. The architecture implemented on Xilinx ISE 8.2 tool indicates that, it is high speed architecture to compress an image and increases the speed by 20% when compared to existing architecture.