Multi-focus image fusion for accurate measurement of nonwoven structures

This paper presents a new region-based image fusion algorithm and its applications for measuring essential parameters of nonwoven structures. The algorithm combines a series of partially focused images of the same sample view captured at different focusing points to form a fully focused image that is fundamental for accurate detections of fiber edges in the structure. It starts with selecting a number of source points based on the maximum gradient matrix, and locating initial fiber boundaries using the pixel-based image fusion algorithm. Within the fiber boundaries, the source points diffuse in the same rate, and the boundaries are formed when their expanding fronts encounter each other. These new boundaries divide the image view into regions of various sizes, each representing a coherent area centered at one source point. Finally, each region is filled with the corresponding region that has the highest average sharpness value among all of the multi-focus images. The paper also presents the experimental results on the fiber diameter, fiber orientation and pore size distributions of nonwovens generated by using this algorithm, in comparison with the results from other methods.