Fig. 1. Lens defocus and focusing model

Fig. 2. Original drawing and corresponding depth drawing. (a)‒(c) Generation of depth maps using defocus cues; (d) depth maps from RGB-D database

Fig. 3. Different defocused images and corresponding depth images in the same scene. (a) Under defocusing; (b) normal defocusing; (c) over defocusing; (d) RGB-D diagram

Fig. 4. Comparison of gradient information between defocus area and focus area. (a) Schematic diagram of clear area and blurred area of defocus image; (b) gradient comparison curves of clear area and blurred area

Fig. 5. Schematic diagrams of Gaussian mixture model with gradient distribution of clear patch and blurred patch. (a) Clear patch; (b) input; (c) blurred patch

Fig. 6. Schematic diagrams of image spectra before and after defocus blur. (a) Original image; (b) spectra of the original image; (c) defocus blur of the original image; (d) spectra of image after defocus blur

Fig. 7. Results of defocusing and clear area separation in pictures with different defocusing degrees. (a) Original input; (b) gradient spike feature separation result; (c) gradient tail feature separation result; (d) image frequency domain feature separation result

Fig. 8. Calculation process of similarity between fused image and truth image. (a) Defocused image; (b) ground-truth image; (c) similarity of fused and ground-truth image

Fig. 9. Binary graph with different similarities and relation graph between index value and similarity. (a)‒(c) Binary images with increasing similarity; (d) relationship between index T₀ and similarity

Fig. 10. Probability diagram of fused image better than original image under different weight intervals

Fig. 11. Ambiguity increasing graph and corresponding characteristic graph in the same scene. (a)‒(e) Defocused image with increasing ambiguity and its corresponding image feature map

Fig. 12. Evaluation values of blurriness degree in the same scene. (a) Normalized value of existing evaluation function (b) T value of this method

Fig. 13. Evaluation value of blurriness degree in different scenarios. (a) Roberts function; (b) DCT function; (c) entropy function; (d) proposed method

Fig. 14. Original drawing and corresponding depth drawing. (a)‒(c) Over defocus images and their generated depth maps; (d)‒(f) under defocus images and their generated depth maps; (g)‒(i) normal defocus images and their corresponding depth maps