Multispectral palmprint recognition is one ofthe most useful biometric techniques due to features obtained from differentspectral resolutions/wavelengths. In this paper, we propose a multidirectionaltransform-based feature encoding plan for reliable and robust representationand matching of multispectral palm images. The method extracts the region ofinterest (ROI) for palmprint images captured with non-contact sensors. Theregistered ROI of each band is newly downsampled using DWT. This approachallows us to take more lines into consideration for interpolation. Aundecimated dual-tree complex wavelet transform based multidirectional featureencoding plan is then newly applied since it provides better shift invarianceand directional selectivity. Finally, abinary code matching strategy with score level fusion is used to computematching for efficient identification. The experimental results obtained onCASIA and PolyU datasets show that the presented method gives better results inthe blurring binary code matching case than state-of-the-art methods andprovides comparable performance in the non-blurring binary code matching.