You are on page 1of 18

Digital Image Inpainting

By Ms. Vidhya M. Shinde Under the Valuable Guidance of Prof. V. U. Deshmukh


1

Seminar on

Literature Survey

Inpainting:-

Introduction

A set of techniques for making undetectable modifications to images


Reconstructing damaged old photographs and removing unwanted objects from images/videos and improving overall composition Automatic Digital Inpainting is a technique which restores damaged image or video by means of image interpolation. Inpainting is the art of restoring lost/selected parts of an image based on the background information in a visually plausible way.

Algorithms:1. Exemplar Based Image Inpainting

2. Successive Elimination Algorithm


3. Multilayer or multi-resolution Image Inpainting Algorithm
4

Exemplar Based Image Inpainting

Figure: Notation Diagram

Computing patch priorities Texture Synthesis Filling order


5

1. Computing patch priorities:priority P(p) as the product of two terms:


P(p) = C(p)D(p) .. (1)

C(p) the confidence term

(2)

D(p) the data term,


.(3)
6

2. Texture Synthesis:
The patch with highest priority is to be computed
Propagate image texture by direct sampling of the source region Search in the source region for that patch which is most similar to p

3. Filling order:-

Match the patch window with entire image from(0,0)th pixel to (x,y)th pixel and identify the matched window

Collect all the matched window position and again find C(p) and D(p) to compute current P(p)
Find highest priority matched window then replace the previous patch window with currently got matched window
8

Successive Elimination Algorithm

To increased the computational efficiency Based on exemplar- based image inpainting algorithm It uses Sum of Absolute Difference (SAD) to obtain global optimal solution

It can be used for Restoration of Small scratches and also for reconstruction of image after removal of large object from the image
9

Steps of SEA algorithm:


1. Sum of Absolute Difference (SAD):

2. Choose the most prior patch 3. Initialize MinSAD :-

4. Compute the Sum of MinSAD


10

Cont.. 5. {Compute the sum of the current patch

6. if ( R-Min SAD M(m,n) R+MinSAD) { compute the SAD value of the current patch; if ( SADMinSAD) { MinSAD=SAD; } } } 7. Find out the position of the patch that the minimum value of SAD corresponds to, and this patch q is the most matching. 8. Copy the information in q p p ; 9. Update the value of C(p) p p ; } 11

Disadvantages

Detects only objects with sharp edges Smoothing may miss edges in presence of noise these algorithms is that the diffusion process introduces some blur e.g.

12

Applications
1. 2. 3. In Photography:Restoration of photographs, films and valuable paintings Removing Red-Eye Zooming and to produce special effects Removal of Occlusion:Removing stamped date, text subtitle, stamp, publicity from images Removing objects to creative effects Removing logos from videos Wireless Image Transmission:To replace lost blocks in coding and transmission of images e.g. in Streaming Video
13 13

Examples
1. Removal of Large object from image:Original image Fill region Inpainted image 50 50 50

100

100

100

150

150

150

200

200

200

250

250

250

300 50 100 150 200

300 50 100 150 200

300 50 100 150 200

14

Examples
2. Reconstruct Noisy image:Original Nature Corrupted - 50% Inpainted Nature

15

Conclusion

16

References

17

18

You might also like