摘要
声呐图像受噪声影响严重,分辨率低,传统算法对其目标分割效果较差,为此,提出了小波域多分辨率MRF模型的声呐图像分割算法。小波域多分辨率分析有利于提取声呐图像弱特征信息;每一分辨率中的观测特征采用高斯混合模型建模,尺度内同标记的观测特征用高斯模型建模,用各向同性的双点多级逻辑(Multi-Level Logistic,MLL)模型建模每一尺度的标记场;最后,用迭代条件模式(Iterated Conditional Mode,ICM)实现多分辨率马尔可夫随机场(Multi-Resolution Markov Random Field,MRA-MRF)中能量函数的最优解,获取标记场,完成声呐图像分割。从视觉效果和定量分析两方面验证。对比实验的结果表明,该文算法能有效地提取声呐图像的弱目标信息,较好地将目标区域和背景区域分割出来,具有较高的分割精度和鲁棒性。
Since sonar image is usually with characteristics of serious noise pollution and low resolution, it is difficult to get object segmentation results with high precision by traditional algorithms. A sonar image segmentation algorithm based on multiresolution Markov rand filed（MRF） model in wavelet domain is proposed. In the wavelet domain, the multiresolution analysis is advantageous to extracting the weak characteristic information of sonar image. Using Gauss mixture model describes the observational characteristics of each scale and the characteristics with a same mark in the intra-scale obey the Gauss distribution. The label field of each scale is modeled by isotropic two-point MLL model. Finally, the optimal solution of the energy function in the MRF model is obtained by using iterated conditional mode（ICM） to get the tag field, and complete the sonar image segmentation. From the two aspects of visual effects and quantitative analysis, to compare the experiment results, the proposed algorithm can extract the weak target information of sonar images effectively, which can better distinguish the target region and the background region, and has higher segmentation accuracy and robustness.
出处
《声学技术》
CSCD
北大核心
2016年第3期198-203,共6页
Technical Acoustics
基金
国家自然科学基金(联合基金)重点项目(U1401252)
国家自然科学基金项目(61272237)
省重点实验室开放基金项目(2015KLA05)资助
