A scalable and programmable architecture for 2-D DWT decoding

The compression of still images by means of the Discrete Wavelet Transform (DWT), adopted in the JPEG-2000 and MPEG-4 standards, is becoming more and more widespread because it yields better performances than other compression methods, such as DCT. The demand of efficient architectures for 2-D DWT coding and decoding for a variety of different applications and embedded systems is rapidly increasing. This paper presents the implementation of a 2-D DWT decoder for Mallat tree decomposition, suitable for low power applications, such as portable devices. The decoder design has been synthesized and validated in 0.35 μm CMOS technology. The architecture is scalable according to the desired maximum image size, the maximum DWT kernel length and arithmetic accuracy, and it is programmable at run-time to process different image sizes and to use different DWT kernels.