基于边缘导向视觉测量图像的LWP-SPIHT精确编码  

Precise Edge-directed LWP-SPIHT Encoding for Visual Measuring Image

作　　者：刘俊杰 杨学友[1] 叶声华[1] 

机构地区：[1]天津大学精密测试技术及仪器国家重点实验室,天津300072

出　　处：《电子与信息学报》2012年第12期2795-2801,共7页Journal of Electronics & Information Technology

基　　金：国家自然科学基金重点项目(50735003)资助课题

摘　　要：为实现对视觉测量图像更加有效地稀疏表示,尽可能地捕捉图像本身的几何特性,该文首先提出一种基于灰度梯度方向预测边缘走向的方法,精确定位图像边缘,并在此基础上运用方向提升格式小波包(LWP)变换与普通提升小波包变换相结合的方式,对整幅测量图像进行分解。其次,采用的各个子带能量信息与方向信息和的代价函数,有效地保持边缘信息的完整性。最后完善并解决了小波包变换在SPIHT(Set Partitioning In Hierarchical Trees)编码中产生的父子节点间的冲突问题。通过对视觉测量中的远、近景图像、局部放大图像以及靶标图像的测试证实,基于边缘导向的提升格式小波包多级树集合分裂(LWP-SPIHT)的视觉测量图像压缩算法,更好地保留了压缩后视觉测量图像的边缘信息,改善了图像的重构质量和精度。In order to achieve sparse representation and capture more geometrical features of visual measuring image, a predicted method of edge information based on gray gradient direction is proposed first. Due to the precise positioning of edge, the method of combination of Lifting Wavelet Packet （LWP） transformation and general wavelet packet transformation is adopted to decompose the whole measuring image. Then the designed cost function maintained the integrity of edge information efficiently, that is the sum of energy mean of every sub-band and direction information. Finally, parent-child conflict of wavelet packet transformation in Set Partitioning In Hierarchical Trees （SPIHT） encoding is completed and r＇esolved. By the test of distant view image, nearby view image, partial enlarged image and target image, the compression method for visual measuring image based on edge-directed LWP transformation SPHT encoding （LWP-SPIHT）, preserved edge information of visual measuring image and enhanced the reconstructed quality and precision.

关 键 词：图像处理 图像编码 灰度梯度方向 方向提升小波包变换 父子节点冲突 

分 类 号：TN919.81[电子电信—通信与信息系统]