Encryption of images has been carried out by chaos-based algorithms using a single sequence for a long time. The existing schemes are not secure enough for batch image encryption. A batch image encryption scheme using a stacked autoencoder network is proposed in this paper wherein two chaotic sequences are used. A shuffling matrix is produced by the first sequence which is used to shuffle image pixels on each input image, and the other sequence generates a series of independent sequences which are used to confuse the permuted images with the encrypted images. In this scheme, parallel computing of the stacked autoencoder reduces the run-time complexity and encryption effect is enhanced. The efficiency of the used scheme is evaluated by a series of analysis techniques. The results and analysis concluded that the proposed scheme is very efficient and is able to resist statistical attack and differential attack. © 2019 IEEE.