Dissertation > Excellent graduate degree dissertation topics show

Research on Blind JPEG Image Forensics

Author: WuShouYang
Tutor: YuNengHai；LiWeiHai
School: University of Science and Technology of China
Course: Information Security
Keywords: Image tampering Blind forensics JPEG compression Quantify the correlation Blockiness
CLC: TP391.41
Type: Master's thesis
Year: 2009
Downloads: 261
Quote: 7
Read: Download Dissertation

Abstract

With the rapid development of Internet technology, the emergence of high-resolution digital camera and a powerful image-editing software, the image to modify the phenomenon more common, although it is to enrich people's daily lives, but to a certain extent, image changes also brought many problems. The tampered image used in the news media or law, will create a big impact on society. Therefore, the authenticity of the digital image identification technology has drawn wide attention and become a current research focus. Currently, the majority of the digital image in the majority of image acquisition device outputs the digital image on the Internet are stored in JPEG format. JPEG image forensics technology has a very important practical value for JPEG images Blind forensics algorithm more constraints there are many gaps from the practical. In this context, this paper carried out the JPEG digital image forensics technology research. The key to the technology of the digital image forensics to solve the problem of blind detection of the authenticity of the images. Thesis research work and contributions are as follows: (1) quantify the correlation JPEG image authenticity blind detection algorithm based on the JPEG compression. This method based on different synthetic image quality factor premise assumption, by detecting the image quality coded quantization process the factor unanimously to detect image been tampered with. The algorithm itself is simple in principle, easy to implement, and the complexity of the algorithm is mainly concentrated in the estimate of the quality factor (q value). Blind detection algorithm proposed by experimental validation, the results show that the algorithm can effectively detect the authenticity of the images, and calibration tampered region. Thesis experimental statistical analysis, summarized when the difference between the value of q of the q value and modify the image of the original image is greater than 5, to detect the correct rate of more than 95%, while when the value of q between the two is too close (less than 5), The correct rate is slightly reduced. (2) the authenticity of the synthesized image based on the blocking effect in JPEG blind detection algorithm. The detection algorithm is based on the JPEG image synthesis, the modified region and the original region at the junction of the 8 × 8 sub-blocks of the dislocation in the frequency domain, the performance of this feature for the attenuation of the original peaks or splitting proposed. Caused by detection during the synthesis, the block does not match, and to determine the authenticity of the region of the synthesized image. First calculate the travel of the power spectrum of the image of each column, and these power spectrum together, and the structure and texture of the image itself to eliminate interference. For a JPEG compressed image, can be found the periodical peak of the power spectrum in the specific frequency, without going through the JPEG compressed image is not the existence of such peaks. The flatness of the mid and high frequency on the local frequency domain energy spectrum for analysis, use of the sliding block of 32 × 64 to detect the horizontal block effect, the sliding block of 64 × 32 detects the vertical block effect, and horizontal and vertical detection The result multiplied to amplify spliced ??region with non-modified regional differences. Experimental results show that the method detects the authenticity of the JPEG synthesized image is very effective. (3) Further analysis of the above two algorithms, given the performance of existing algorithms to compare and judge its performance merits. The blockiness inconsistency blind image detection algorithm based on comparing the proposed algorithm with the existing literature, can be drawn proposed to quantify the correlation image blind detection algorithm based on JPEG compression in terms of accuracy, or in terms of Immunity are better than the algorithm proposed in the literature; blockiness of JPEG-based synthetic image blind detection algorithm proposed in this paper is basically the same algorithm in Immunity literature, but judged on selected slider tampered region. its performance is slightly inferior to that of the algorithm proposed in the literature.

Related Dissertations

CLC: > Industrial Technology > Automation technology,computer technology > Computing technology,computer technology > Computer applications > Information processing (information processing) > Pattern Recognition and devices > Image recognition device
© 2012 www.DissertationTopic.Net Mobile