Reversible data hiding in encrypted image with separable data extraction from image decryption

We propose a method of reversible data hiding in encrypted image. Proposed method achieves reserving room before encryption, and separates data extraction from image decryption. Our method is an improvement of Ma's method [IEEE Trans Inf Forensic Secur 8(3): 554-558, 2013]. Our improvements mainly focus on two aspects. (1) We improved the interpolation error estimate method via using Bicubic interpolation instead of pixel estimation that calculates the weighted sum of four surrounding pixels. Thus more sharp interpolation error histogram is obtained to increase the hidden information capacity. (2) We use partitioned local histogram shift instead of traditional histogram shift to reduce the amount of shifted pixels. This directly results in a higher quality of stego image in the same embedding capacity. The experimental results indicate that the improved method offers better performance. It is superior to Ma's work in both embedding rate and the PSNR values of stego images.