摘要
在分析了JPEG2000感兴趣区域(ROI)编码标准的基础上,提出了一种基于形态学的动态JPEG2000的ROI提取方法.利用形态滤波器组与分水岭分割算法将LLn小波子带图像划分为若干闭合区域,再利用区域纹理特征来判定LLn小波子带图像的ROI以及生成LLn子带的二值模板,最后依据离散小波变换结构以及坐标映射原理,动态生成全部子带的二进制ROI模板.在JPEG2000压缩软件中的应用结果证明,所提方法克服了目前ROI编码必须人工参与的缺陷,利用它可根据图像内容自动生成ROI编码图像,大大提高了批量ROI编码的效率.该方法同时还支持任意形状的图像,并与国际JPEG2000标准兼容.
After analyzing the region-of-interest （ROI） coding standard in JPEG2000, a novel dynamic JPEG2000 ROI extract method based on morphology was presented. The morphologic filter group and watershed segmentation algorithm in the wavelet LLn subband image were used to obtain several closed regions. By using local texture features the little regions were merged into large ones. After the ROI of the LLn image was ascertained based on the regions＇ area and location information, the ROI binary template of LLn image were formed. Finally the ROI binary template of LLn was used to construct the whole wavelet image ROI template which includes all subbands according to the DWT （discrete wavelet transformation） structure and the wavelet coordinate mapping theory. This method can realize the ROI coding of JPEG2000 based on the image content automatically so that it can greatly improve the efficiency of ROI coding. It supports the arbitrary shape ROI and is compatible with the international standard of JPEG2000, and its feasibility and validity have been proven in a JPEG2000 compression software.
出处
《西安交通大学学报》
EI
CAS
CSCD
北大核心
2006年第2期125-128,共4页
Journal of Xi'an Jiaotong University
基金
国家高技术发展计划资助项目(2002AA135150)
国家自然科学基金资助项目(60021302)
教育部科学技术研究重点资助项目(104168)
