Multichannel audio coding is studied from a rate-distortion theoretical viewpoint. Two practical coding techniques, both of which are based on rate-distortion optimization, are also proposed. The first technique decorrelates a multichannel signal hierarchically using elementary unitary transforms. The second method rearranges a multichannel signal into sub-signals and compresses them at optimized bit rates using a conventional codec. Both objective and subjective tests were conducted to illustrate the efficiency of the methods.