Fig 1.
An illustration of 3D object imaging with an optical camera.
IA and IB are image pixels of point A and B respectively. With the current focal length, the surface that point A lies on is focused while that of point B is not. It can be seen1 that in-focus pixels in the image plane form a continuous region. By adjusting the object distance to the lens, a series of defocused (part-in-focus) images could be obtained.
Fig 2.
Flowchart of multi-focus images fusion based on RMLP approach.
Fig 3.
An illustration of density-connectivity with same mask label.
Fig 4.
Illustration of RMLP with two multi-focus images and three pyramid layers.
Fig 5.
Illustration of the S1 Dataset and its fusion results.
(a)-(f) are six random sampled examples from fifty multi-focus images, (g) is the mask image with only EOF measurement, (h) is the mask image with the proposed DBRG segmentation algorithm. (i) is the fusion result of the Laplacian pyramid (LP) method [11] and (j)is the fusion result of the proposed RMLP.
Fig 6.
Illustration of the S2 Dataset and its fusion results.
(a)-(f) are six random sampled examples from sixty multi-focus images, (g) is the mask image with only EOF measurement, (h) is the mask image with the proposed DBRG segmentation algorithm. (i) is the fusion result of the Laplacian pyramid (LP) method and (j)is the fusion result of the proposed RMLP.
Fig 7.
Illustration of quantitative evaluation (a) RMSE and (b) SSIM under different DBRG radius for the focus region mask segmentation. (c) RMSE and (d) SSIM with different pyramid layers. (e) and (f) are comparisons of nine methods.
Fig 8.
Average execution time of RMLP and other methods.