多特征驱动的彩色多聚焦图像融合理论与方法研究

For the reason that the selection of color space model limits the feature measure and extraction of focused regions, the fused image cannot describe the pixel feature of the focused regions, which affects the result of image analysis and understanding. Color multi-focus image fusion has been one of the most important issues that must be solved in image fusion. Based on color multi-focus image fusion and its related key point of the best color space model, this project studies theories and methods of multi-feature-driven color multi-focus image fusion. In order to improve the quality of fused image, the project proposes a color space model for color multi-focus image fusion based on the impact of various color space on fusion process. In order to improve the accuracy of focus measure, the project proposes a new model of focus measure based on deep learning. In order to improve the validity of focused region extraction, the project proposes a model of focused region extraction based on visual attention. Starting from the construction of color space, this project performs the color multi-focus image fusion and provides the theoretical basis for color image fusion and multichannel image processing. The expected achievements have important application value in the field of smarter cities and military operations.

由于颜色空间模型的选择限制了聚焦区域的特性度量和提取，融合图像不能准确描述聚焦区域像素特征，影响了图像的准确分析与理解，彩色多聚焦图像融合已成为图像融合领域亟待解决的问题之一。针对此问题，本项目以彩色多聚焦图像融合为基本背景，以构造适合彩色多聚焦图像融合的最优颜色空间模型为切入点，研究多特征驱动的彩色多聚焦图像融合理论与方法。根据不同颜色空间对融合过程的影响，提出适用于彩色多聚焦图像融合的颜色空间模型，以提高融合图像质量。针对彩色多聚焦图像的聚焦区域特性判定问题，提出基于深度学习的聚焦区域特性判定模型，以提高聚焦特性判定的准确性。针对彩色多聚焦图像的聚焦区域提取问题，提出基于视觉注意的聚焦区域提取模型，以提高聚焦区域提取的有效性。本项目从构造颜色空间出发，完成彩色多聚焦图像融合任务，为彩色图像融合及多通道图像处理提供新的理论依据，预期成果在智慧城市和军事作战等领域具有重要应用价值。

颜色空间模型的选择对彩色多聚焦图像融合有重要影响，现有颜色空间不能准确描述聚焦区域像素特征，限制了聚焦区域的特性度量和提取，导致对图像场景不能准确分析与理解。本项目以彩色多聚焦图像融合为基本背景，以构造适合彩色多聚焦图像融合的最优颜色空间模型为切入点，以提高彩色多聚焦图像融合质量为目标，研究了多特征驱动的彩色多聚焦图像融合理论与方法。研究了适用于彩色多聚焦图像融合的颜色空间模型构建方法；研究了利用不同尺度变换工具实现的图像分解方法，研究了图像特征识别与分类方法，给出了多特征驱动的彩色多聚焦图像聚焦区域特性判定模型；研究了颜色空间下彩色多聚焦图像融合方法，提高了融合图像质量。项目完成了计划书规定的研究内容，实现了预期研究目标。本项目已正式发表期刊类学术论文13篇，其中SCI收录论文7篇，EI收录论文2篇，中文核心期刊论文4篇（以上数据未去重），出版学术专著1部，授权中国发明专利2项、实用新型专利1项，申请国家发明专利4项。本项目以颜色空间构建为切入点，完成彩色多聚焦图像融合任务，为彩色多聚焦图像融合及多通道图像处理提供了新的思路和理论依据。