Research On Image Fusion Algorithm Based On Compressive Sensing

Image fusion is a technology that combines the information of the same scene of two ormore pictures from identical or different types of sensor to generate a more precise,comprehensive and reliable image description or interpretation of that scene, which is moresuitable for the purposes of human visual perception or computer-processing tasks such assegmentation, feature extraction and target recognition. It has been applied to targetmonitoring, airport navigation, security surveillance, intelligent transportation, medicalimaging and diagnosis, machine vision, geographic information system, intelligentmanufacturing, industrial process and military and etc. In recent years, with the researchprogress of compressed sensing theory, image fusion algorithm based on compressed sensingdomain is gradually got the attention of people. However, the present study also mainly focuson the Discrete Fourier Basis or Discrete Cosine Basis as sparse base under the condition ofthe fusion method, while in reality most of image and video signals are coded by wavelettransform (JPEG2000), therefore, a further research on image fusion theories and algorithmsbased on compressive sensing domain used wavelet as sparse base will have more realisticeffect. The main work of this paper is as follows:(1) In this paper, on the basis of reading a large number of domestic and foreignliteratures, the paper reviews the background, significance and present situation of imagefusion technology, compressed sensing and image fusion based on compressed sensingdomain and image fusion quality evaluation.(2)The paper introduces theoretical framework of compressed sensing and principle ofmain steps’ algorithms what are sparse representation, constraints and structure ofmeasurement matrix and reconstruction algorithm. Then the paper compares the advantagesand disadvantages of the various methods.(3)The paper proposes an image fusion algorithm based on compressed sensing whichused the wavelet as sparse base. Firstly, the algorithm decomposes the original images bywavelet transform. As low frequency coefficients of wavelet transform are not sparse, thealgorithm directs sampling all low frequency part. While to the high frequency coefficients,they are sparse and different decomposing layers and directions of high frequency coefficientsare different, so the sparse degree is calculated and then the number of measurement of eachpart is computed. Secondly, fuse the low frequency part by the method that combines regionalgradient and energy. Fuse the measurements of high frequency parts by using the method ofthe largest absolute. Thirdly, the algorithm reconstructs the high frequency part by OMP CSreconstruction method. Finally, the fused image is reconstructed by inverse wavelet transform. The result of experiment shows that the objective and subjective evaluation is effective. Andcompared to the traditional algorithm, our algorithm can not only save the cost oftransmission, storage, storage and reconstruct, but also the result of fusion is good.(4)The paper proposes an image fusion quality assessment algorithm based oncompressed sensing domain. It is a new image fusion evaluation method which combinescompressed sensing with structural similarity. Firstly, after the fusion in CS measurementsdomain, and before reconstruction, it evaluates the measurements by the structure similarityfusion evaluation method to contrast the change of the measurements value before and afterfusion which side evaluate the effect of the fused image. Experiments results show that, to acertain extent, evaluating the quality of fused image in CS domain is effective and efficient.(5)In the end, the paper summarizes the main work of this paper, and describes someproblems that in the study, and look to the future research direction.