如何利用显著性分析高效、准确地完成遥感影像感兴趣区域提取是遥感影像处理领域亟待解决的关键问题之一，对有效缓解遥感影像高速获取与低速解译间的矛盾具有重要意义。针对现有方法仅考虑单幅影像显著性，在背景复杂或有噪声干扰时无法准确提取感兴趣区域的问题，本项目拟提出具有较强抗噪能力的联合显著性分析模型，首先从具有相似地物特征的多幅遥感影像入手，构建共性显著特征分析模型，排除显著性较高的背景干扰；其次整合单幅影像自身的光谱、纹理与边缘特征，建立内部显著特征分析模型，使感兴趣区域显著性进一步增强；再次基于对数共生直方图设计具有抗噪能力的统计显著特征分析模型；最后融合各显著特征图，通过分割完成多幅遥感影像感兴趣区域的准确提取。此外，项目还将结合几何特征评价指标实现对遥感影像显著性分析的有效评估。上述研究将为准确提取遥感影像地物目标开辟新的思路，其成果也能为遥感影像压缩及变化检测等相关方向提供必要技术支持。

How can saliency analysis be utilized in the efficient and accurate detection of a region of interest (ROI) has become one of the key problems in the field of remote sensing image processing. It has great significance in effectively alleviating the contradiction between high-speed access to the remote sensing images and low-speed data interpretation. The extant models generally cannot precisely extract ROIs against complicate background or noise interference by only relying on single-image saliency. In this project, we propose a joint saliency analysis model with strong anti-noise ability. Firstly, inter-saliency analysis model is built to effectively depress salient background by considering multiple remote sensing images with similar ground object features. Secondly, intra-saliency analysis model is built to further enhance the saliency of ROIs by integrating spectral, texture and edge features within a single image. Thirdly, stat-saliency analysis model with anti-noise ability is constructed based on logarithmic co-occurrence histograms. Finally, all the saliency maps from the above models are fused to achieve accurate multiple-image ROI extraction after threshold segmentation. In addition, the project will incorporate a set of geometric feature evaluation indices for effective accuracy assessment in the saliency analysis for remote sensing images. In sum, this research work will offer a new idea for the accurate extraction of ground objects in the remote sensing images. Moreover, its results will provide necessary technical support to the remote sensing image compression, change detection and other related fields.