|
|
|
题名
|
作者
|
年代
|
出处
|
被引量
|
| 1 | 基于游程和扩展指数哥伦布编码的任意形状感兴趣区域图像编码显示文摘给出一种上下文自适应的游程编码和扩展指数哥伦布编码。利用游程编码算法对图像小波系数及ROI掩模进行上下文自适应建模并输出三元组样本;然后扩展普通的指数哥伦布编码,使其可以编码由游程编码建模输出的三元组样本,在对小波系数编码的同时可以携带感兴趣区域掩模标记信息。由此得到一种可以区别感兴趣区域和背景区域的高效编码算法,并以此算法为基础提出一种感兴趣区域编码的编解码框架,该框架包括5/3小波变换、小波域掩模标记生成、不均匀最佳量化、游程编码和扩展的指数哥伦布编码。该算法的游程建模过程简单,熵编码算法可用闭合公式表达,具有较高的可实现性。实验结果表明,提出的算法支持多个任意形状的感兴趣区域,感兴趣区域相对于背景区域的编码优先级可调,并且可以获得高于基于BbB-shift的SPIHT算法的压缩性能。 | 徐勇 徐智勇 张启衡 | 2011 | 光学精密工程2011，19，1： | 9 |
| 2 | 嵌入掩膜的SPIHT任意形状ROI编码显示文摘感兴趣区域(ROI)编码可以在低码率条件下获得高质量的局部感兴趣区域,或在图像渐进传输中使感兴趣区域获得优先传输。本文在分析了当前各类ROI编码方法的基础上,基于SPIHT算法提出了一种支持多个任意形状感兴趣区域并生成可任意截断码流的ROI编码算法。该算法在SPIHT算法中嵌入了重要系数的ROI掩膜信息,使编码器同步地进行图像和ROI形状的编码,使得生成的码流具有任意可截断的特性。文中还就图像ROI编码的质量评价指标进行了讨论,并给出了一种充分考虑ROI和背景的重要性与面积比例差别的图像质量评价指标,称为重要性-面积加权峰值信噪比(WPSNR)。实验结果表明,该算法支持有损到无损的多个任意形状ROI的图像编码,而且ROI优先级可调,能够生成具有嵌入式可截断性质的码流,在任意地方截断仍能保证解码器所需的图像信息和ROI掩膜信息,且计算复杂度和SPIHT相当,压缩效果高于BbB移位算法。适用于低码率应用或感兴趣优先渐进传输的应用。 | 徐勇 徐智勇 张启衡 左颢睿 | 2009 | 光电工程2009，36，9： | 4 |
| 3 | 无链表图像感兴趣区域编码算法显示文摘针对基于链表实现的感兴趣区域编码算法占用存储资源较多的问题,提出了一种无链表的编码算法.在SPIHT(等级树集合分裂)编码过程中,采用标志位图表示系数和集合的重要性信息;优先编码感兴趣区域,利用队列缓存非感兴趣区域系数和集合信息;编码非感兴趣区域时,从队列中恢复编码所需的重要性信息.编码过程不需要提升感兴趣区域小波系数,能实现感兴趣区域重建质量的精确控制.仿真实验表明,该算法优于提升小波系数的感兴趣区域编码算法;当编码码率为1 bpp(比特/像素)时,其存储需求仅为链表实现的感兴趣区域分离编码算法的1/10. | 潘波 杨根庆 孙宁 | 2010 | 西南交通大学学报2010，45，1： | 3 |
| 4 | 一种基于感兴趣区域元数据的视频转码方案显示文摘提出一种面向客户端带宽与设备受限的视频数据转码方案。首先,获取并存放与视频资源相关的用户感兴趣区域(ROI)元数据。然后,在视频传输阶段,根据客户端的带宽、处理器处理能力、功耗的不同情况,参考ROI信息,将原始视频进行转码。转码后的视频流仍会保持较好的主观观看效果。 | 赵新峰 王恒 | 2012 | 计算机与现代化2012，，9： | 0 |
| 5 | Image compression algorithm of floral canopy based on mask hybrid coding for ROI显示文摘To achieve high-quality image compression of a floral canopy,a region of interest(ROI)mask of the wavelet domain was generated through the automatic identification of the canopy ROI and lifting the bit-plane of the ROI to obtain priority of coding for the ROI-set partitioning in hierarchical trees(ROI-SPIHT)coding.The embedded zerotree wavelet(EZW)coding was conducted for the background(BG)region of the image and a relatively more low-frequency wavelet coefficient was obtained using a relatively small amount of coding.Through the weighing factor r of the ROI coding amount,the proportion of the ROI and BG coding amount was dynamically adjusted to generate embedded,truncatable bit streams.Despite the location of truncation,the image information and ROI mask information required by the decoder can be guaranteed to achieve high-quality compression and reconstruction of the image ROI.The results indicated that under the same bit rate,the larger the r value is,the larger the peak-signal-to-noise ratio(PSNR)for the ROI reconstructed image and the smaller the PSNR for the BG reconstructed image.In the range of 0.07-1.09 bpp,the PSNR of the ROI reconstructed image was 42.65%higher on average than that of the BG reconstructed image,43.95%higher on average than that of the composite image of the ROI and BG(ALL),and 16.84%higher on average than that of the standard SPIHT reconstructed image.Additionally,the mean square error of the quality evaluation index and similarity for the ROI reconstructed image were both better than those for the BG,ALL,and standard SPIHT reconstructed images.The texture distortion of the ALL image was smaller than that of the SPIHT reconstructed image,indicating that the image compression algorithm based on the mask hybrid coding for ROI(ROI-MHC)is capable of improving the reconstruction quality of an ROI image.When the weighing factor r is a fixed value,as the proportion of ROI(a)increases,the quality of ROI image reconstruction gradually decreases.Therefore,upon the application of the ROI-MHC image compression algorithm,high-quality reconstruction of the ROI image can be achieved through dynamically configuring r according to a.Under the same bit rate,the quality of the ROI-MHC image compression is higher than that of current compression algorithms of same classes and offers promising application opportunities. | Sun Guoxiang Wang Xiaochan Ding Yongqian Li Yuhua Zhang Baohua Li Yongbo Zhang Yu | 2017 | International Journal of Agricultural and Biological Engineering2017，10，5： | 0 |
您所在的机构暂未开通此部分的功能，请联系管理员