微型电脑应用

针对多聚焦图像的特点，采用非下采样剪切波变换(NSST)对源图像进行多尺度分解，从而获得低频系数和高频系数。对于低频系数，采用改进的空间频率作为脉冲耦合神经网络(PCNN)输入，基于PCNN加权规则进行融合；对于高频系数，使用区域能量、区域方差和绝对值取大的多特征规则进行融合。经NSST逆变换重构图像。实验表明，所提算法生成的融合图像边缘清晰、细节丰富，可以得到较好的视觉效果和较优的评价指标。

In view of the characteristics of multi-focus images, non-subsampled shearlet transform(NSST) is adopted to perform multi-scale decomposition on the source image, thereby obtaining low frequency and high frequency coefficients. For the low frequency coefficients, the modified spatial frequency is adopted as the input of pulse coupled neural network(PCNN), and the fusion is based on the PCNN weighting rule. For the high frequency coefficients, the multi-feature rules of regional energy, regional variance and the selection of maximum absolute values are used for fusion. The image is reconstructed by NSST inverse transform. Experiments show that the fused image generated by the proposed algorithm has clear edges and rich details, and obtains better visual effects and better evaluation indicators.