To address the security and fraud issues associated with transmitting sensitive plaintext image data over public channels, a secure multi-party hybrid encryption-sharing scheme, which integrates the 2D sine-cosine (2D-SC) chaotic system and P-tensor product compressed sensing (PTP-CS), is proposed. Firstly, the 2D-SC chaotic system is presented, and its state variables and performance are thoroughly analyzed. Secondly, by combining the 2D-SC chaotic system with the Logistic chaotic system, a sensing matrix and keystream are constructed. Under the control of the keystream, sparse processing, position permutation, compressed sensing, and bidirectional diffusion are applied to the plaintext image, generating a secret image and effectively reducing the correlation between pixels. Subsequently, the secret image is embedded into a non-sensitive carrier image using image steganography, creating a cipher image that provides dual protection for both the content and visual aspects of sensitive plaintext data. Finally, the cipher image is segmented into N shadow images and transmitted to N recipients through a public channel. Only when at least r recipients (r represents the minimum number of shares required for the reconstruction of the image, r < N) decide to initiate the image recovery process, the image can be successfully restored. The experimental findings validate that the proposed encryption scheme guarantees the secure recovery of sensitive data while resisting various common attacks, and exhibits strong visual security and compression performance.