Research On Steganography Of High Dynamic Range Image

Since the birth of the Internet in 1969,it has made contributions that cannot be ignored in various fields,and it has also brought great challenges to the field of information security.Information hiding technology is developed in response to information security issues,using different forms of communication media to achieve the transmission of secret information,one of the branches is steganography.Digital imagebased steganography can be divided into two categories: spatial domain and transform domain.It has gradually developed from non-adaptive steganography that treats all pixels indiscriminately to adaptive steganography that focuses on image content combined with encoding.With the increasing popularity of the field of deep learning and the increase of computer computing power,the use of deep learning network models for training to directly generate steganographic images and the use of confrontational ideas for steganalysis have also become a promising development direction for today's image steganography technology.However,most of the existing researches use LDR images as steganographic carriers,and there are certain limitations in the trade-off between steganographic security and steganographic capacity.Compared with traditional images,high dynamic range images have a great dynamic range,because the data is floating-point number,so as to truly display various details such as color and brightness in natural scenes.Therefore,they have received more and more attention in the field of digital images.At the same time,the floating-point number data type also makes the redundancy space of HDR images larger,which also provides new ideas for image steganography.This thesis conducts a detailed study on different storage formats of HDR images,based on traditional image steganography and deep learning image steganography,two different steganography algorithms using HDR images as carrier images are proposed:1)The HDR image steganography algorithm based on RGBE format combines the idea of adaptive steganography algorithm to divide the image into flat areas and edge areas,and the pixel modification is determined according to the index value of the pixel point of the HDR image.The amplitude can be modified to realize the embedding of secret information.It has been verified that the perceivable probability of the steganographic image generated by this algorithm is very low,and it can reach the SCOA security level.Compared with the existing steganography algorithm based on HDR image,the steganographic image quality is better.2)The HDR image steganography algorithm based on the convolutional network uses the Open EXR format to extract the last eight bits of the HDR image Mantissa space as the steganography carrier.The hidden network based on the Residual Block and the extraction network are designed,and the the cosine similarity function is added to the loss function to measure the color distortion between image pairs,finally realizes that an LDR image is hidden in the same size HDR image.The quality evaluation of the generated images from both subjective vision and objective indicators has achieved good results.In addition,it can be seen from the experimental results that this algorithm has high security and can resist attacks from steganalysis tools.