Research On Reversible Data Hiding With High Dynamic Range Images

Today,digital watermarking is an important mean of protecting the security,integrity,and copyright of data.The traditional watermarking technology will lead to the permanent destruction of the cover media,and many valuable data,such as military,medical or forensic data,are not allowed to be permanently destroyed.Therefore,reversible data hiding(RDH)technology has emerged,which enables the original cover media to be perfectly restored after the watermark is extracted from the cover.In recent years,the application of high dynamic range images has become more and more extensive.This technology enables pictures to display the details of the pictures more finely,especially those with scenes with very strong contrast.High dynamic range images come in many formats.The method in this thesis is mainly based on one of the special and most widely used formats,and the Radiance RGBE format.The image in this format adds an exponential channel e compared to the traditional image,which affects the color of the pixel exponentially,so we also call it the luma channel.This thesis uses the characteristics of e-channel to conduct reversible data hiding research.The main research contents are as follows:1.Reversible data hiding based on Reverse-Golomb Code in high dynamic range imageWhen the Radiance RGBE format is used for reversible data hiding,the embedding of messages can cause pixel correlation damage,causing an effect which called stream expansion.This effect causes the storage space of the image to increase after the message is embedded.In order to solve this problem,we analyzed the pixel characteristics of the image,and came to the conclusion that the cover in this scene can be regarded as an all-zero cover.So we designed a special encoding method named Reverse-Golomb Code.The code can encode the message into sparse coding,so that the alteration of the pixel is greatly reduced in the embedding process of the message,thereby greatly reducing the stream expansion.Our coding has a slightly higher coding efficiency than the most popular STC code,and has a much higher coding rate.At the same time,our coding is not only applicable to the scenarios in this thesis,but also to all all-zero covers.2.Reversible data hiding based on inconsistent distortion model in high dynamic range imageSince most scenes displayed in high dynamic range images have strong contrasts,it is unreasonable to use a consistent distortion model for reversible data hiding in these images.Therefore,we introduce the latest model of reversible data hiding,the inconsistent distortion model,to solve this problem.The so-called inconsistent distortion is to define the distortion for the pixel according to the pixel value and the surrounding pixels.In this thesis,we use the characteristic of e-channel values in Radiance RGBE format to define the distortion of message embedding,and then design an data hiding framework based on this distortion.The experimental results show that our method has a certain improvement in image quality compared with the previous methods.