改进区域分割的多聚焦图像融合方法

Multi-focus Image Fusion Method Based on Improved Regional Segmentation

针对传统多尺度几何分析方法在图像融合时易损失清晰度以及融合规则选取复杂的缺点,提出一种改进基于区域分割的多聚焦图像融合方法;首先,根据提升静态小波变换快速获得初始融合图像,并对初始融合后的图像进行Normalized Cut算法处理以获得不同的分割区域,然后分别对原始图像进行NSCT变换并计算每个分割区域内高频分量系数的绝对值之和,最后选取绝对值最大的区域为融合后区域,并通过遍历每个分割区域获得融合后的多聚焦图像;数值试验证明,本算法不但融合规则选取较传统多尺度分析融合算法简单,还能够有效克服清晰度损失的缺点,具有有效性。

According to the shortcoming that traditional multi-scale geometry analysis method is easy to lose clarity and hard to choose fusion rules when an image is fusing,this paper proposes an improved multi-focus image fusion method based on regional segmentation,firstly,uses Lifting Stationary Wavelet Transform to obtain the initial fused images,then receives different segmented regions by treating the initial fused image with Normalized Cut algorithm,uses NSCT transform for the original images respectively,calculates the sum of absolute value of transform coefficient of each segment of high frequency component,finally chooses the biggest absolute value area for fused area,and obtains fused multi-focus image by iterating through each segmented region. Numerical example proves that this algorithm is not only more simple in choosing fusion rules than traditional multi-scale analysis fusion algorithm but also can overcome the shortcoming of clarity loss and has validity.