An introduction to computational crystallography:the relationship between aluminum-based spinel structures and their morphologies

Computing technology is developing very fast and has permeated among many fields of sci-ence and technology, giving rise to many interdisciplinary subjects. Crystallography, originating from the activity of mining and smelting, has now become a comprehensive subject concerning crystal growth mechanism, chemical composition, structural morphology and physical properties and its research scope is widened by introducing many new achievements from basic sciences such as physics, chemistry, and math…     

自适应PCNN多聚焦图像融合

脉冲耦合神经网络（PNNN）型参数多,计算复杂,并且通常选取经验常数作为链接强度,这极大地限制了PCNN的普遍适用性.针对该问题,在分析PCNN模型基本特征的基础上,结合多聚焦图像的基本特点提出了一种基于PCNN的自适应多聚焦图像融合新方法.该方法在PCNN简化模型的基础上既将拉普拉斯能量作为PCNN对应神经元的链接强度β,又将其作为PCNN对应神经元的反馈输入经过PCNN点火从而获得每幅参与融合图像的点火映射图,最后通过选取适当的融合规则获得融合图像.实验结果表明了该方法的有效性,这种有效性不仅体现在视觉效果上,而且体现在客观评价标准上.     

一种基于PCNN的多聚焦图像融合改进算法

为提高图像融合质量和融合效率,提出一种基于脉冲耦合神经网络的多聚焦图像融合改进算法。对待融合的源图像作分块处理,选取合理的图像质量评价指标,计算每个分块的指标值,归一化后相减得到指标差值。把指标差值作为外部刺激输入到PCNN模型中,得到脉冲输出结果。用脉冲输出与给定的阈值作比较,若输出脉冲超过阈值则选择指标值大的源图像块作为融合图像块,否则取指标值小的源图像块。选取互信息、交叉熵、均方根误差、峰值信噪比、结构相似度以及相关系数6个客观质量评价指标进行评价,实验结果表明,该算法可获得较好的图像融合效果。