05-05-2012, 04:11 PM
Simultaneous Structure and Texture Image Inpainting
simultaneous.pdf (Size: 1.12 MB / Downloads: 36)
INTRODUCTION
THE filling-in of missing information is a very important
topic in image processing, with applications including
image coding and wireless image transmission (e.g., recovering
lost blocks), special effects (e.g., removal of objects), and
image restoration (e.g., scratch removal). The basic idea behind
the algorithms that have been proposed in the literature is
to fill-in these regions with available information from their
surroundings. This information can be automatically detected
as in [5], [10], or hinted by the user as in more classical texture
filling techniques [8], [14], [28].
IMAGE DECOMPOSITION
In this section, we review the image decomposition approach
proposed in [30], [31], which is one of the three key ingredients
of the simultaneous texture and structure image inpainting
algorithm. As explained in the introduction, this decomposition
produces images that are very well suited for the image
inpainting and texture synthesis techniques described in the next
sections. The description below is adapted from [31], where
the technique was first introduced. The interested readers are
referred to this work for more details, examples, and theoretical
results.
IMAGE INPAINTING
We now describe the third key component of our proposed
scheme, the algorithm used to fill-in the region of missing information
in the bounded variation image . For the examples in
this paper we use the technique developed in [5]. Other image
inpainting algorithms such as [2], [3] could be tested for this
application as well. The key idea behind these algorithms is to
propagate the available image information into the region to be
inpainted, information that comes from the hole’s boundary and
is propagated in the direction of minimal change (isophotes).We
should also note that these works explicitly showed the need for
high order partial differential equations for image inpainting (in
order to smoothly propagate both gray values on gradient directions),
thereby making simpler denoising algorithms such as
anisotropic diffusion not appropriate.
CONCLUSIONS AND FUTURE DIRECTIONS
In this paper, we have shown the combination of image
decomposition with image inpainting and texture synthesis.
The basic idea is to first decompose the image into the sum
of two functions, one that can be efficiently reconstructed
via inpainting and one that can be efficiently reconstructed
via texture synthesis. This permits the simultaneous use of
these reconstruction techniques in the image domain they
were designed for. In contrast with previous approaches, both
image inpainting and texture synthesis are applied to the region
of missing information, only that they are applied not to the
original image representation but to the images obtained from
the decomposition. The obtained results outperform those
obtained when only one of the reconstruction algorithms is
applied to each image region.