An Improved Multi-focus Image Fusion Algorithm For Wireless Sensor Networks

Wireless sensor networks(WSNs)as one of the hot spots can be applied in the special environments to sample,process and transmit information.As a kind of novel information acquisition and processing technology,it plays a more and more important role in real world.But sensor nodes in WSNs are generally powered by small capacity batteries,their capability of process,storage and communication is weak,which limits the application of the networks.The data fusion of wireless sensor networks is the process to integrate data from multiple sensors with the purpose to decrease redundancy in the networks.It can save more energy and capacity,and prolongs network lifetime.Therefore,it's necessary to research data fusion technique in WSNs.This paper is accomplished through the research of numerous references.Firstly,the background and research status of image fusion in wireless sensor networks is elaborated to demonstrate the significance of multi-focus image fusion in WSNs.Then,the current multi-focus fusion methods are concluded and classified and some typical algorithms are introduced in detail.On account of the shortages of existing fusion algorithms,an improved multi-focus image fusion method which applies the quad-tree structure to realize the adaptive block size with a novel focus measure is presented.The main work of the paper is as follows:Firstly,a novel focus measure of the weighted region-based energy of gradient is proposed.In the process of multi-focus image fusion,it's very important to detect the focus regions.So the majority focus measures used in the references are listed,such as variance,energy of gradient,and energy of Laplacian,etc.And the performance of the proposed focus measure is compared with them by experiments.Then,the proposed algorithm based on quad-tree structure and the novel focus measure is described in detail.It mainly includes two parts:the decomposition strategy according to quad-tree structure and the reconstruction of fused image.The threshold of image decomposition in the quad-tree structure is designed based on weighted region-based energy of gradient to obtain focused blocks of the source images.For the isolated small blocks and holes in focus region of the decision map,consistency check is adopted according to the correlation of adjacent pixels.Finally,the effect of proposed algorithm and the other five algorithms are tested with several groups of multi-focus images,and evaluated in subjective aspects via fusion images as well as difference maps.Objective evaluation is finished via the indexes of information entropy,mutual information,structural similarity and edge preservation.The results show that the proposed algorithm could acquire higher objective indicators and more excellent visual effect,which proves that the proposed algorithm is an effective method to realize multi-focus image fusion.We also take time consumption into consideration,and compare it with fusion method based on DCT in video sensor networks to prove the feasibility of the proposed method for wireless sensor networks.The energy consumption of fusion node is simulated before and after the fusion method,which conforms that reduce of transmission data can save network energy.