Research On Multi-focus Image Fusion Based On Activity Level Measurement

With the development of information industry technology,many image sensor devices have been produced to meet people's needs in a variety of fields.However,due to the limitations of some imaging mechanisms,using only one imaging mechanism often is not enough.For example,in a camera imaging device,the information collected by different focus points varies greatly in the same scene.Under the single condition of the imaging device,the rich multi-source information in the same scene cannot be obtained.Therefore,in the same scene,how to integrate images to get a clear image becomes an urgent research topic according to different imaging principles or parameter.Image fusion is a technology that combines multiple images into one.The fused image contains complementary information from each original image.Image fusion technology is widely used in photography,surveillance,medical,remote sensing technology,aviation,computer vision and other fields.so image fusion research is of great significance.According to the above background and significance of image fusion research,this thesis conducts in-depth research on multi-focus image fusion.Traditional methods based on multi-scale analysis tend to cause information loss due to sampling methods or fusion rules.In the sparse representation method,the dictionary is often insufficient,to obtain the details,and the calculation is expensive.In the early stage,the space-based method is easy to cause block effect.For the above problems,multi-focus image fusion method based on fractional differential and based on Gaussian Laplacian are proposed.The main work and innovations of this paper are as follows:? In the traditional multi-scale analysis methods,image information loss is easily caused by sampling and fusion strategies.In the sparse representation method,the dictionary expression ability is often insufficient,resulting in blur details in fused image and very high fusion time complexity.For the multi-focus image fusion method based on the spatial domain,the algorithm for measuring the image activity is very critical.Fractional differential features are proposed to measure the activity of themulti-focus images.The algorithm first convolves the image with a fractional template in eight directions,and then accumulates the absolute value after convolution in each direction to obtain the activity degree metric of the original image.Each metric map is then compared separately using a sliding window technique.The window with a higher summation value is regarded as the focus,and the corresponding score map is incremented by one.The decision graph is obtained by the score map information.Finally,the final fused image is obtained by weighting the original image by the decision graph.The algorithm effectively overcomes the traditional space-based block effect and so on.Through experimental comparison and analysis,the experimental results of this algorithm are better in both subjective and objective evaluation.? A multi-focus image fusion algorithm based on Gaussian Laplacian(log)is proposed for the problem of easy loss of information and time complexity.Firstly,the Gaussian Laplacian mask is used to filter the original image,and its absolute value is obtained to obtain the focus degree metric of the corresponding original image.Then,using the sliding window technique,each metric map is compared separately,and the inside of the window and the large value are regarded as the focus,and the corresponding score map is incremented by one.The decision graph is obtained by the score graph and a certain strategy.Finally,the original image is weighted by the decision graph to obtain the fused image.Experimental results were performed on a multi-focus image open database.The proposed method is superior to the comparison algorithm in subjective detail and several objective evaluation indicator.and effectively overcome the shortcomings including detail distortion and high time complexity of traditional methods such as sparse representation.