感兴趣区域图像压缩与可见水印技术

【作者】 朱媛媛；

【导师】 张小红；

【作者基本信息】 江西理工大学 ， 通信与信息系统， 2010， 硕士

【摘要】 随着多媒体技术应用的快速发展,图像数据量日益庞大,需要大的存储容量和宽的传输信道,由于图像内部相邻像素之间存在大量的冗余信息,使得图像压缩得以实现,并成为当前的研究热点。为适应各种场合的应用需求,人们对当前的图像压缩系统提出了更高的要求,不仅要求它具有高的压缩性能,还需要增加更多新的功能。小波变换以及嵌入式编码已经成为当今图像压缩的主要技术,越来越受到人们的重视。为了提高图像编解码后感兴趣区域恢复的质量以及它的安全方面的性能,本文对彩色图像分割、感兴趣区域的提取、彩色图像压缩、感兴趣区域编码和可见水印技术进行了一些较有成效的研究,主要成果有:⑴通过研究彩色空间变换后各色彩通道能量分布情况和传统嵌入式编码的量化方法,设计了一种基于新量化方案的彩色图像压缩算法。该方案根据图像小波分解后系数分布特点,结合SPIHT(Set Partitioning in Hierarchical Trees Algorithm)算法的基本原理,实现对各分量小波系数的统一编码。与原始量化方案相比,该方案实现的图像压缩性能更优,尤其是在较低码率下,重构图像能达到更好的压缩效果。⑵结合人眼显著性特性和形态学分水岭算法,设计了一种彩色图像分割及感兴趣区域提取算法。该方法有效地改善了形态学分水岭算法的过分割现象,实现图像的良好分割,在背景简单的图像中能将感兴趣区域准确提取出来。⑶将ROI提取方法与感兴趣区域图像压缩编码相结合,实验结果表明,在较低的压缩比下,恢复的图像中感兴趣部分能保持较好的质量。⑷为更好地保护图像的版权信息,设计了一种感兴趣区域的自适应可见水印方案。水印的嵌入深度根据宿主图像纹理特性而定,在小波域内将可见水印嵌入在图像的感兴趣区域内,并将可见水印的嵌入过程融合在图像的编码过程中,结果表明,可见水印能半透明的呈现在宿主图像感兴趣区域内,使得感兴趣区域得到更好的保护,实现实时性高、安全性好、压缩比不减小的图像数字水印方案。基于以上四个方面的研究,可见水印自适应嵌入在图像的感兴趣区域内,且与彩色图像压缩过程同步实现,这不仅提高了彩色图像的压缩性能,且使图像的版权信息得到了更好的保护。更多还原

【Abstract】 With rapid development of multimedia technology, the amount of image information became larger and larger, so it needs large storage capacity and a wide-band transportation channel. Image compression is achievable because of the redundant information between the adjacent pixels in an image, now it has become a hotspot. In order to satisfy all kinds of application requirement in different environment, the standard of image compression system has been improved, it not only requires a higher performance on compression, but also needs other new functions.At present, in the aspect of image compression, embedded coding algorithm based on wavelet transform is the key technology, and increasingly importance has been attached to it. In order to improve the quality and safty of ROI after image coding and decoding process, some research on image segmentation has been done, ROI extraction, such as color image compression, ROI coding and visual watermark technology, the achievements can be concluded as follows:⑴By studying the energy distribution of the color channels in the transformed color space, and the quantitative method in the traditional embeded coding, a new quantitative method for color image compression was designed, which is based on the distributing of wavelet coefficients. Combined with the keystone of SPIHT (set partitioning in hierarchical trees algorithm), the wavelet coefficients in every color channel are quantized and coded in unity. In comparison with the original quantitative scheme, the new scheme gives better image compression performance, especially in the lower bit rate, the reconstructed image after impression has higher quality.⑵Combined with the significant characteristic of human eyes and morphological watershed algorithm, this paper designed a new segmentation region-of-interest extraction algorithm used for the color image. After the segmentation process, the image retained a very fine edge of the targets, over-segmentation caused by watershed algorithm has been greatly improved, and ROI can be exactly extracted from simple background.⑶The ROI extraction method can be applied in ROI coding, in large compression rate, the experiment result shows that the quality of ROI in reconstructed image is higher than BG (background).⑷An adaptive visual watermark technical in ROI was designed, which can better protect the copyright of the image. The embedding depth is fixed by every pixel’s texture feature of host image. In wavelet domain, the visual watermark is embedded into the region of interest in the image adatively, and this step can be set during the image coding. Experimental results have shown that watermarked image meet the requirements of visible watermark quite well, watermark is translucently showing in the image and ROI does receive better protection, what’s more, it realizes the watermark timely, safty and no effect on compression rate.Based on the above studies, visual watermark adaptive embedded in the objects of the host image, which is realized synchronously with compression. It not only improves the performance of color image compression, but also better protects the image’s copyright.更多还原