基于像素提取的多聚焦图像融合及仿真实验

Multi-Focus Image Fusion by Pixel Extraction and Simulation

根据实际多聚焦图像的特点,摒弃了变换系数融合法和分块提取法的融合思想,以提取清晰像素点为目标,提出了一种全新的基于像素提取的多聚焦图像融合方法。该方法以源图像中清晰的边缘、轮廓特征为提取依据,物理意义明晰,融合方式合理。通过对像素提取率和正确率辩证关系的详细分析,给出了针对一般多聚焦图像融合的提取测度邻域大小和提取阈值的合理范围。经基于该方法最简单策略的融合仿真实验结果表明:该方法是一种有效的多聚焦图像融合方法,能有效克服其他方法所产生的人工副效应和块效应现象,融合结果的主、客观评价指标均有很大提升,具有很好的继续研究价值和应用前景。

According to real multi-focus images, a novel fusion method is proposed by pixel extraction. Rather than adopting the transformed coefficients fusion method or the sub-blocks fusion method, the method is aimed at extracting sharp pixels. It takes the characteristics of the clear edges and contours in the source images as the extraction basis, so it has a more precise physical meaning and its operations are more reasonable. After analyzing the relationship between the pixel extraction ratio and the correct extraction ratio, the ranges of the extraction condition＇s neighbor size and the extraction threshold are given. Comparing the experimental results obtained by using the simplest algorithm based on this method with the ones using algorithms from other methods, it is concluded that the proposed method, which is free from the undesirable artificial effects and block effects, is most efficient in terms of visual effect and objective evaluation. The proposed method is therefore more promising in the fields of both research and application.