In this brief, we show that by exploiting the inherent symmetry of the discrete wavelet transform (DWT) algorithm and consequently storing only the nonrepetitive combinations of filter coefficients, the size of required memory can be significantly reduced. Subsequently, a memory-efficient architecture for DWT/inverse DWT is proposed. It occupies 6.5-mm2 silicon area and consumes 46.8- μW power at 1 MHz for 1.2 V using 0.13- μm standard cell technology. © 2009 IEEE.