Multi-focus Image Fusion Based On Imaging Fuzzy Mechanism

When we observe an area, each object in the scene has its own distance from the observer. So it is hard to get every object showed clearly in one picture when forming images by optical sensors. Image fusion technical helps solving the problem mentioned above. This method extracts the clear parts from different pictures which have different focuses, forming one image with the advantage that every object is showed clearly on it. Since image fusion is the previous stage for further processing, this paper paid attention to the various strategies of it and proposed improved fusion algorithms. The details are listed as follows:Through the research of how optical system forming image, we analyzed the main factors which lead to the image fuzzy:the overlaid circle of confusion and the low-pass filter of focus. Build up the relationship between fussy degree and depth of field. Find a reasonable expression of fussy degree or clearness. Construct a vector to evaluate the degree of clearness for the extraction from original images. By a simple multi-focus image fusion algorithm based on image-forming principle, we examined the correctness of our evaluation vector. Then through a fusion method based on double-regulations, taking correlation as the standard to judge the clearness level, finally get a fusion image by choosing the clear part from different original images. Show the validity of our proposal.After the research of image fusion algorithms in different domains, a Laplace pyramid image fusion algorithm based on the measurement of pixel-wise correlation has been given. Apply Laplace pyramid operation to every original picture. Take different fusion methods according to high frequency part and low frequency part. Then reverse the Laplace pyramid operation to get the final fusion image. Experiments showed that our proposal has an improvement comparing with others.