BWI Watermarking 설계에 근거한 칼러 영상의 손상검출과 복구에 관한 연구

Abstract

멀티미디어 사회를 마지하여 영상물의 복원 및 소유권에 대한 논쟁이 일어 날 수 있어 지금 까지 소유권 주장에 대한 연구가 되어 왔다. 그러나 소유권과 부가적으로 손실이 된 경우 원 영상의 복원 요구되어 이에 대한 연구가 필요하다. 따라서 본 논문은 소유권과 손실의 있다면 복원을 할 수 있는 방법에 대한 연구로 소유권,손상부분 검출 및 복원을 BWI 워터마킹 설계의 기반에서 연구한 것이다. 논문의 논제,구성등의 타당성 및 창의성,그리고 재현성에 대해서 심사하였다.|For the information era, there exists a kind of technology that is information hiding. However, in that era, information hiding techniques received much less attention from the research community and from industry than cryptography. Until since the growth of computer power and digital era, people start to be concerned about protecting copyright because malicious attackers usually try to modify meaningful information of an image to change its meaning. Therefore, in order to solve this problem, steganography and watermarking has been presented. Steganography is a method that establishes a covered information channel in point-to-point connections, whereas watermarking does not necessarily hide the fact of secret transmission of information from third persons. For the watermarking, it requires a certain perceptual threshold allowing the insertion of additional information to adjust between perceptible degradation and imperceptible degradation to fulfill visible or invisible watermarking. So, in this sense, watermarking represents a specific application of steganographic techniques.

In this thesis, we propose a digital watermarking scheme about color image of tamper detection and recovery using block-based watermarking. Firstly, in our watermark embedding, each color channel (R, G, and B) is divided into 2×2 non-overlapping blocks. The embedding space is created by setting the three LSBs (least significant bits) of selected each block to zero, which will hold the authentication and recovery codes. The authentication codes are obtained through doing exclusive-or (XOR) both the extracted local features of image blocks and the generated global features of the original color image. The recovery codes are generated using block-mapping original color image and using color model conversion. Secondly, in the tamper detection, instead of examining each embedded authentication data, we get the global features of the tampered watermarked image through doing XOR both all the extracted authentication codes and the generated local features of tampered watermarked image blocks. And then we utilize a majority-selecting technique to detect the legitimacy of image blocks. Lastly, in the tamper recovery, we propose 3 stages to accomplish recovery ability. Experiment demonstrates that our scheme is oblivious, correctly localizes the tampering wherever the tampered positions are and can be successfully accomplish to recover with acceptable visual quality.