留学生理工科dissertation-图像融合和增强方法

Abstract

图像融合是一个重要的整合成一个紧凑的形式的相干时间和空间信息的可视化技术。拉普拉斯算子融合的流<br/>程的来自不同来源的图像到一个单一的融合图像，根据每个区域的一个显着选择规则，结合地区。在本文中，我们提出了丹算法的方法，使用口罩金字塔，以更好地定位选择过程。一个面具金字塔在不同尺度的图像，以提高融合图像质量超出一个全球性的选择规则。该掩模金字塔方法的几个例子中所提供展示其性能在各种应用程式时新的嵌入式系统架构，该架构基于阿卡迪亚II视觉处理器提出。

 Image fusion is an important visualizationtechnique of integrating coherent spatial and temporalinformation into a compact form. Laplacian fusion is aprocess that combines regions of images from differentsources into a single fused image based on a salienceselection rule for each region. In this paper, we proposedan algorithmic approach using a mask pyramid to betterlocalize the selection process. A mask pyramid operatesin different scales of the image to improve the fusedimage quality beyond a global selection rule. Severalexamples of this mask pyramid method are provided todemonstrate its performance in a variety of applications.A new embedded system architecture that builds uponthe Acadia  II Vision Processor is proposed.

关键词：图像融合，融合曝光，高动态范围成像，聚焦融合，图像融合，降噪，图像增强。

Keywords: image fusion, exposure fusion, high dynamicrange imaging, focus fusion, image blending, noisereduction, image enhancement.

1 Introduction
Since Burt and Adelson published their first paper onLaplacian pyramid-based image processing in 1983[1-4],the multispectral and multiscale representation of animages has been used extensively in image compression,stabilization and fusion. The pyramid is essentially a datastructure consisting of a series of band-pass filtered copiesof an image, each representing pattern information on adifferent scale. The image fusion, represented by theLaplacian pyramid of each source image, extracts localsalience information from each source image at multiplepyramid levels ranging from coarse to fine, and thenreverses the pyramidal operation to form a fused image.The extraction of salient features is based on a specificselection rule. One common selection rule compares thelocal pixel energy of each pyramid level and selects thepyramid level with the maximum absolute value for eachlocation. Other rules include globally weighted averagingof the Laplacian images of a pyramid level.[5] TheLaplacian fusion scheme has been mainly used for imageintegration of the mixed modality sensors.[6] It is alsoused in image stitching and multi-focus fusion for mono-modal sensors.[7]Sarnoff’s Acadia I vision processor was the firstembedded solution for real-time fusion, capable ofaligning and fusing two NTSC/PAL images in real-time[8].

The second generation Acadia II visionprocessor can align and fuse three 1280x1024 images inreal time, and is also capable of globally and locallyenhancing the source images. This local enhancement isdone per pyramid level in the Laplacian images thatnormalize local contrast. Additionally, Waterfall SolutionsLtd. has put forward an FPGA based real-time imagefusion hardware device. The hardware supports simpleweighted averaging as well as high performance multi-resolution Laplacian fusion.[9]The use of image fusion as a primary vision processingfunction has gained widespread acceptance over the pastseveral decades. Presently, both hardware and softwaresolutions exist to fuse individual pixels in order to providean integrated view of a scene captured via multiple sensorsor sensor conditions (e.g. focus, exposure). The fusionfunction can provide increase informational content to theviewer when a priori information about the source imagesis made available. Thus the purpose of this paper is not todiscuss the higher level fusion schemes, but rather topresent methods of extending current fusion capabilitiesand applications for real time embedded systems. Presentfusion schemes typically only apply a global selection ruleto all image pixels. In many applications, it would bepreferable to use a localized rule. Therefore, we proposethe generation of a mask pyramid in the hardwarearchitecture to localize pixel selection. We will presentexamples of this in this paper, demonstrating a genericarchitectural implementation for the fusion systemdesigned to consider. If any intelligent analysis from thesource imagery needs to be used in fusion, #p#分页标题#e#http://www.ukthesis.org/dissertation_sample/ it can beprogrammed into this mask pyramid. We will alsodemonstrate that with this new capability, we can easilyextend the fusion application to image enhancement, highdynamic range compression and image blending amongother applications.
The paper is organized as follows: the basic andadvanced Laplacian fusion model are discussed in SectionII. Section III presents examples of how the mask pyramidcan be used to extend the fusion function to perform imageenhancement, depth-of-field extension, high dynamicrange compression and image blending. Section IVaddresses the system architecture and Section V concludesthe discussion.

5 Discussion and Conclusion
This paper has discussed the use of the mask pyramid toextend the conventional pyramid based fusion architectureto other image processing applications. The innovationcomes from the use of local information to determinesalience and thus influence the feature selection map. Thecanonical implementation is configured not to just executesingle applications (as discussed in Section III), but isflexible enough to combine multiple applications in acomprehensible way. For example, the hardware can beconfigured to simultaneous perform both DoF and HDRCprocessing while locally enhancing the source image.[13]The hierarchical local-information-encoded maskingtechnique presented herein is not limited to pyramidfusion, but may apply to any wavelet fusion. The inputdata source can be multi-dimensional, and is not limited tovision data. The feature selection rule as described canoutput local weights when the register is enabled, and isthus not confined to binary.

References
[1] P. J. Burt and E. H. Adelson, “The Laplacian pyramidas a compact image code”, IEEE Trans. Commun. COM-31, 532-540, 1983.
[2] E. Adelson, C.H. Anderson, J.R. Bergen, P.J. Burt andJ.M. Ogden, “Pyramid methods in image processing”,RCA Engineer 29, 33-41, 1984.
[3] P. J. Burt, “The pyramid as a structure for efficientcomputation”, Multi-resolution Image Processing andAnalysis, A. Rosenfeld, ed., Springer-Verlag, Berlin, 1984.
[4] J.M. Ogden, E.H. Adelson , J R. Bergen and P.J. Burt,“http://www.ukthesis.org/ Pyramid-based computer graphics”, RCA Engineer, 30-5,1985.
[5] P. J. Burt, and R. J. Kolczynski, “Enhanced imagecapture through fusion”, Proc. International Conferenceon Computer Vision, Berlin, pp. 173-182, 1993.
[6] Moira I. Smith, Jamie P. Heather, “Review of imagefusion technology in 2005”, Thermosense XXVII,Proceedings of the SPIE, 5782, pp. 29-45, 2005.