Lossy Compression With An Adaptive Parameters Set On Pixel Encrypted Images

In recent years, with the rapid development of Internet technology, Internetapplication moves closer to the homes. More and more people like to use socialnetworking and cloud computing. When people enjoy the convenient, theseapplications also brought some hidden trouble.The user often need to send theirpictures to network provider. When the receiver want to get the sender’s image,hedoes not directly get the images from the sender, but to download on a webserver.This convenience is obvious, but still pose a safety hazard.First of all, thenetwork service provider may be not reliable, he leak may images that containpersonal privacy;Second, malicious eavesdropper can also download user imagesfrom the network provider on the other malicious purpose.As a result, users want toperform an encryption to protect his privacy before send the image to the networkprovider, but a lot of network terminal such as mobile phone can’t afford thecompressing. For network provider, when he received the encrypted images fromthe user, he is eager to compress the encrypted image by considering the factorssuch as bandwidth and storage space.There are many encryption algorithm forcompressed images.But the research on compression for image in encrypted domainis relatively lack, especially compress the image with the pixel values aremasked.In order to solve this problem, this paper proposes an image compressionscheme with an adaptive parameters in encrypted domain. The user performs apreprocessing to decompose the original image into four group according the texturelevels.Put the blocks with the same texture degree into one group.So thequantization step can be different in different group.The user also encrypts the pixelvalues with a standard steam cipher.The network provider decomposes the encryptedsubimage with different texture degree into a downsampled subimage and severaldata sets with a multilple-resolution construction.An encoder quantizes thesubimage and the Hadamard coefficients of each data set to reduce the data amountwith different quantization step int different group.At the receiver side,while asubimage is decrypted to provide the rough information of the original image,thequantized coefficients can be used to reconstruct the detailed content.In this paper, the main contributions of the proposed method are listed below:(1) A texture degree based compression in encrypted domain is proposed.Thescheme can be divided into: preprocessing,image encryption,compression inencrypted domain,image reconstruction.In the preprocessing,we decompose theoriginal image into four group depend on the texture degree.The different group hasdifferent quantization step.Compared with the whole image has one quantization step,the advantage is obvious. The images have their pixel value encrypted with astandard stream cipher.The reasons of why we choose the standard stream cipher areshown in this paper.When we perform compression,the encrypted groups aredecomposed into into a downsampled subimage and several data sets. An encoderquantizes the subimage and the Hadamard coefficients of each data set to reduce thedata amount with different quantization step int different group.(2) We apply an adaptive parameters set with the scheme propose above.Thegroups in different images can adjust quantization step.The proposed method canselect the best system parameters according to the smoothness of the block.