In the era of big data, protecting digital images from cyberattacks during network transmission is of utmost importance. While various image encryption algorithms have been developed, some remain vulnerable to specific cyber threats. This paper presents an enhanced version of the image encryption algorithm based on bit-plane extraction (BPCPD) to address its vulnerability to chosen-plaintext attacks. The proposed cryptographic system encompasses three primary phases. The initial phase involves bit-plane extraction from the plaintext image and the generation of random sequences and a random image using multiple chaotic maps, such as the chaotic Arnold map and the chaotic CAT map. The second phase is dedicated to permutation operations, which comprise three sub-phases: multi-layer permutation, multi-round permutation, and recursive permutation. In the third phase, diffusion is introduced to the permuted image through pixel substitution, coupled with XOR operations performed on the respective bit-planes of the random image. To gauge the efficiency of the proposed encryption scheme, a range of experimental analyses are conducted, including histogram analysis, contrast assessment, entropy measurement, correlation analysis, encryption quality assessment, and investigations into noise attacks and occlusion attacks. The results of these experimental analyses, in comparison to an existing encryption scheme, demonstrate that the proposed framework surpasses both BPCPD and other existing encryption schemes in various aspects of performance.