The existing reversible data hiding in encrypted image (RDHEI) algorithm has the problems of poor utilization rate of pixels, and small embedding capacity. Addressing such issues, an RDHEI algorithm based on adjacency difference and block classification (ADBC) scheme is proposed. Firstly, the spatial correlation of the original image is fully utilised to obtain the adjacency difference image by calculation, and the initial block classification operation is realised according to the maximum value of the image block; secondly, the median edge predictor is used to predict the pixels for the non-embeddable image blocks in the initial classification to complete the second block classification; then, the image encryption is executed; finally, the auxiliary data and secret data are embedded in the encrypted image by bit substitution. The experimental results show that the average embedding rate (ER) of algorithm in this paper is over 0.06 bpp,0.01 bpp and 0.15 bpp higher on the BOSS base, BOWS-2 and UCID datasets respectively compared with the existing algorithms.