【作者】 姚燕；

【导师】 罗斌； 汤进；

【作者基本信息】 安徽大学 ， 计算机应用技术， 2012， 硕士

【摘要】 随着数字多媒体技术的不断发展和网络技术的普及,数字图像在电脑、数码摄像机、照相机、数字电视、智能手机等数字产品中已经得到了广泛的应用,图像以数字形式存储、传输和共享已经是时代发展的必然。在数字形式下,图像可以被滤波、变换或与其他图像合成,而且图像在以数字形式进行传输的过程中,往往会有恶意的修改致使图像内容在传输过程中发生改变,例如,移去图像中特定的物体、修改图像的内容、在图像中添加一些违法的标志信息等等,我们称这种不经允许而修改图像内容的操作为图像的篡改。针对上述情况,为了还原图像的内容,就诞生了数字图像修复技术,它可以很好的还原被篡改的数字图像。数字图像修复的原理是利用受损图像中的已知信息,然后按照特定的方法来进行修补,使得修复后的图像尽量接近或等于原图像。本文将详细的介绍两种基于分形编码的数字图像修复算法,一种是基于迭代分形解码和边缘扩展的数字图像修复算法；另一种是基于双份分形码解码与自适应TV算法的图像恢复算法。由分形编码的原理可知,分形解码本质上就是分形编码的一个逆过程,它是通过使用原始图像编码生成的分形信息和一幅与原图尺寸相同的初始图像来生成一个吸引子,该吸引子是原始图像的一个近似。在进行分形解码时,初始图像是任意选取的一幅图像,然后利用迭代函数进行解码,使得解码后的图像近似原始图像。由于分形解码使用的初始图像是任意选取的,所以通常它与原始图像的相似性是很小,而在图像修复过程中,我们希望初始图像与原始图像尽量的相似,所以本文介绍了第一种利用分形编码进行图像修复方法,即基于迭代分形解码和边缘扩展的数字图像修复方法,该方法利用修复后的图像作为初始图像再次迭代解码来修复图像,使得初始图像在迭代的过程中越来越接近原图像,从而使分形解码后得到的吸引子越来越近似或等于原始图像。同时,为了确保图像块与其存储分形信息块之间最大程度的不同时被篡改,算法在水印嵌入阶段使用了两次推操作和换操作,使获得的索引表能更好的将分形信息的存储位置分散开。为了更好的修复大面积篡改的图像同时也能保证修复后的效果很好,本文还介绍了另一种利用双份分形码的图像修复算法,即基于双份分形码解码与自适应TV算法的图像修复算法。该算法在图像中嵌入了两份分形码,一份是原图中主要信息的分形编码,一份分形码是原图轮廓信息的编码。为了保证这两份分形信息不被同时被篡改,算法对两份分形信息的提取和存储都做了处理。在分形码解码时,首先利用接收到的图像中的轮廓信息的有效分形码进行分形解码、利用自适应TV算法进行图像修复,恢复得到原图像的主要结构信息,其次是利用提取详细信息的分形码解码来对上步修复的图像进行二次恢复。算法首先对原始图像进行采样生成采样图像,然后分别对原始图像和采样图像进行分形编码。对采样图像的分形码块,从它的最中间开始,从最中间开始以循环的形式由内向外的提取,对原图的分形码也是用相同的方法提取。组合这两个分形码,得到包涵原图轮廓信息和细节信息的分形码,然后依据生成的索引表将分形码存储到原始图像中。最后将Wong水印嵌入到图像LSB1位上,获得嵌入双份分形码和Wong水印的图像。在图像恢复阶段,算法首先利用了Wong水印算法来定位图像被篡改位置,然后先利用嵌入的采样图像的分形信息和自适应TV算法进行迭代恢复生成初级修复的图像,最后再利用原始图像的分形信息进行迭代修复。论文在第三章和第四章的后半部分分别给出了两种算法的修复实验和效果对比,来证明这两种修复算法的有效性。更多还原

【Abstract】 With the progress and development of the digital technology and network, the digital image technology has been widely used in computer network, digital cameras, digital cameras, digital televisions and mobile phone. Therefore, images in digital form for storage, transmission and sharing are the inevitable achievements of technology development. The image can be filtering, transform and synthesis with other images when stored them in digital form, so, sometimes there will be a lot of factors which will defect the information of the image for some purpose, when the image is transform in digital form, such as remove specific objects of the image, modify the image contents, add some extra information to original image and so on. To deal with the above problems, the digital image restoration techniques emerged. Digital image inpainting technology refers to use the known information of the image and a specific method to repair the damaged image region, such that the restored image is as close as the original image. The main content of this thesis is the use of fractal coding method to digital image restoration.This thesis introduces two kinds of digital image restoration algorithm by fractal coding. The first restoration algorithm is based on the iterative fractal decoding and edge extension, the second restoration algorithm is based on the double fractal code decoding and adaptive TV algorithm. The principle of Fractal Decoding is using a initial image and the fractal information of original image to approximate the original image. In fractal decoding, we use an initial image and the iterative function to generate a decoded image to approximation the original image. In the image restoration process, we hope that the initial is very similar to the original image. But the initial image fractal decoding is selected randomly, so the similarity between the initial and the original image is often very small. The method based on the iterative fractal decoding and edge extended made an improvement to the usually digital image restoration algorithm base on fractal code, it uses the restored image as the initial image for the next iterative fractal decoding and image restoration. In order to ensure the image block and the fractal storage block not be tampered together, the algorithm use two push operations and an exchange operation to obtain the index table which can dispersed storage location as soon as possible. At the same time, this algorithm only needs the last two significant bits to store information, which can improve the signal to noise ratio of the embedded image.Another image restoration algorithm is based on double fractal code decoding and adaptive TV algorithm. In fractal decoding, first of all, we use the received fractal code about the contour information and adaptive TV algorithm to restore the main structure of the original image, followed by the use of fractal code decoding to extract detailed information secondary recovery steps to repair the image. Then, we use another fractal code which contains the detail about the original image to repair the tampered image again. The process between the first fractal decoding repair and the second process is basically the same, the difference is that, in the first iterative decoding, the initial image for fractal decoding is selected randomly; however, the initial image in the second part of the decoding is generated by the first part of image repair. During the next iteration, the initial image of the first part of the fractal decoding is the image generated by the last iteration of repair. Firstly, the original image is sampled to generate a sample image, and then extract the fractal code of the original image and the sampling image. We first obtain the fractal code of the middle position of the sampling image use the loop form from inside to outside; because we default think that the most important content of the image is in the middle, so as the fractal code of the original image. Combine these two sub-codes; we can get a fractal code which both contain the contour and details information of the original image. And then according to the generated index table, we embed the fractal code into the original image. Finally, embedding the Wong watermark into the LSB1bits of the image to obtain the watermark image. In image recovery, algorithm first use the Wong watermark to locate the tamped region, and then iterative use extracted fractal code of the sampling image and the adaptive TV algorithm to restore the tampered image to generate a primary repair image which contain the contour information of the original image, finally, we iterative use fractal information which contains the details of the original image for the last repair. This thesis proposed two kinds of image restoration algorithm through a lot of experiments, and the experimental results show that the two image restoration algorithms are validate.更多还原