This paper presents an improvement of the data sorting method for reversible data hiding in QR code images by applying an EVM technique to sort the prediction error values before embedding. The order of the predicted error values was recalculated and is replaced the original local variance, which was calculated from the four neighboring pixels to get the maximum possible PSNR value. The new variance can extend the level of the pixel area to a maximum of 30 levels with 40 versions of the QR code images that is used for testing. The results showed a more efficient sorting over previous methods, especially small data embedding.
Keywords
prediction error; reversible data hiding; QR code; EVM