This paper proposes a wrapping based directional Hartley transform (WDirHT) in the Hartley domain. The wrapping based curvelet transform is used to construct the WDirHT in the Hartley domain. The WDirHT is a sparse representation than curvelet transform. The redundancy factor of the transform is 2.82 with computational complexity of O(N2log2N) for NxN image and takes less computation time for reconstruction. The WDirHT is used for dontent based image retrieval (CBIR) application. The sparse nature of WDirHT coefficients is exploited to derive the feature vector for CBIR application. The CBIR algorithm has been tested on different image database and the results show that the retrieval rate is better compared to the several multiresolution CBIR methods. © 2008 IEEE.