<p>This work introduces a unique approach to multi-focal image fusion (MFIF) involving the concepts of both transform and spatial domains. The tool used for image decomposition in the transform domain is Hilbert vibration decomposition (HVD). HVD is an adaptive signal decomposition tool that splits an image into a set of instantaneous energy components and provides better spatial resolution. These energy components generally consist of instantaneous amplitudes and frequencies in which amplitudes are used for fusion. For this, the quadtree method of the spatial domain is used to generate the optimal size of blocks of the generated instantaneous amplitudes of the source images. Then, focused amplitude and defocused amplitude blocks are predicted in this structure with the help of the sum of weighted modified Laplacian (SWML). After this, morphological and small region filters are used to avoid any blur/unwanted lines in the focused part. Finally, fused energy components are generated by the fusion rule, and all these fused energy components are summed up to develop an all-in-one focus image. It is seen that by using this strategy, the output image is in better focus, and the problem of limited capturing of an imaging sensor in the camera is also reduced.</p>

错误:搜索内容不能为空,请输入英文关键词
错误:关键词超出字数限制,请精简
高级检索

A Hybrid Multi-Focal Image Fusion Scheme Based on Hilbert Vibration Decomposition and Quadtree Structure

  • Gaurav Choudhary,
  • Dinesh Sethi

摘要

This work introduces a unique approach to multi-focal image fusion (MFIF) involving the concepts of both transform and spatial domains. The tool used for image decomposition in the transform domain is Hilbert vibration decomposition (HVD). HVD is an adaptive signal decomposition tool that splits an image into a set of instantaneous energy components and provides better spatial resolution. These energy components generally consist of instantaneous amplitudes and frequencies in which amplitudes are used for fusion. For this, the quadtree method of the spatial domain is used to generate the optimal size of blocks of the generated instantaneous amplitudes of the source images. Then, focused amplitude and defocused amplitude blocks are predicted in this structure with the help of the sum of weighted modified Laplacian (SWML). After this, morphological and small region filters are used to avoid any blur/unwanted lines in the focused part. Finally, fused energy components are generated by the fusion rule, and all these fused energy components are summed up to develop an all-in-one focus image. It is seen that by using this strategy, the output image is in better focus, and the problem of limited capturing of an imaging sensor in the camera is also reduced.