SUMMARYThis paper introduces a reversible watermarking algorithm that exploits an adaptable predictor and sorting parameter customized for each image and each payload. Our proposed method relies on a wellknown prediction-error expansion (PEE) technique. Using small PE values and a harmonious PE sorting parameter greatly decreases image distortion. In order to exploit adaptable tools, Gaussian weight predictor and expanded variance mean (EVM) are used as parameters in this work. A genetic algorithm is also introduced to optimize all parameters and produce the best results possible. Our results show an improvement in image quality when compared with previous conventional works. key words: prediction-error expansion (PEE), histogram shifting (HS), reversible watermarking, Gaussian weight predictor, expanded variance mean (EVM)