Inpainting
![](https://upload.wikimedia.org/wikipedia/commons/thumb/f/f8/Restoration.jpg/350px-Restoration.jpg)
Inpainting is a method of image processing that attempts to reconstruct destroyed or lost parts of an image. The goal is to create a reworked image that the viewer cannot tell that changes have been made.
application areas
Reconstruction of missing parts of the picture
Old pictures can have cracks or individual parts can be completely missing. With the help of inpainting, the missing parts of the image can be reconstructed.
Remove objects
When removing objects or people from an image, the only user input required is to mark the object to be removed. The object is then automatically replaced. The main difficulty is to draw the background as steadily as possible in terms of color and texture into the area to be filled.
Image compression
Inpainting can also be used as image compression. In order to achieve compression, blocks of pixels are deleted. A relatively large number of blocks are deleted in areas of constant texture and only a few along edges, so that the edge course is not lost. After the deleted blocks have been reconstructed, there is no discernible loss of quality compared to JPEG compression.
origin
All procedures that deal with image inpainting are based on the craft of reconstructors. This is an old profession mainly practiced by picture restorers. The process called inpainting is the successive continuation of the edge information into the inside of the picture (filling a gap). Reconstructors proceed as follows:
- The entirety of the picture gives an impression of how the picture should be continued. The result should be a coherent picture.
- The contour lines that touch the edge of the mask are continued into the image along their imaginary extension, as is the structure of the area around the mask.
- The contour lines in the mask define different regions and are filled with colors that match the edge.
- The texture is added, i.e. H. small details like white dots on a blue uniform sky are painted to make the picture look natural.
Most of the digital algorithms also use this procedure to some extent.
Algorithms
Structural inpainting focuses on the reconstruction of the structure and edges in the missing part of the image. One of the most important structural inpainting methods is PDE. The algorithm iteratively extends lines of equal gray values ( isophotic lines ) at the edge into the area to be filled. Then paint is drawn into the area of the image to be filled. There are also a few other approaches such as B. the Navier-Stokes Fluid Flow method .
While Textural Inpainting only concentrates on the reconstruction of the texture, with Combined Structural and Textural Inpainting structures and textures can be reconstructed.
With image-based rendering procedures are for image synthesis in which the geometric modeling phase is replaced by an image recording process referred. Since all information of a three-dimensional object can never be reproduced on images, an inpainting step is introduced which reconstructs missing information as precisely as possible.
swell
- ↑ Object Removal by Exemplar-Based Inpainting ( Memento of October 12, 2012 in the Internet Archive ), Criminisi, A, Perez, P., & Toyama, K., Appears in the Proceedings of the 2003 IEEE Computer Society Conference on Computer Vision and Pattern Recognition ( PDF )
- ↑ M. Bertalmío, G. Sapiro, V. Caselles and C. Ballester., "Image Inpainting", Proceedings of SIGGRAPH 2000, New Orleans, USA, July 2000.