Generating higher dynamic range scene via fusion integration based on DWT and SVM

To expand the dynamic range of intensity levels in actual scenes, high dynamic range imaging is needed which is able to distinguish detail information in the scenes accurately throughout highlight areas and shadow areas at the same time. Digital image fusion is the potential solution to capture critical information from scenes covering both lightest and darkest areas. Wavelet analysis is one of leading techniques for image fusion, which decomposes digital images by means of a set of basis functions. At each level of 2D discrete wavelet transform (DWT), the source image is decomposed into four images of a quarter size at a coarser scale, resulting in one approximation coefficient and three detail coefficients. Two source images can be merged by simple fusion operations such as averaging, minimizing and maximizing. To further enhance image fusion quality across a broader dynamic range, support vector machine (SVM) is proposed which is used to extract a well-trained hyperplane for binary decision making in the fusion process. This integration approach demonstrates the efficiency and effectiveness on data retrieval. Visual appealing depicts remarkable improvement accordingly. In addition, the quantitative metrics of output fusion images have demonstrated the advantages of fusion integration.