DOIONLINE

DOIONLINE NO - IJASEAT-IRAJ-DOIONLINE-2985

Publish In
International Journal of Advances in Science, Engineering and Technology(IJASEAT)-IJASEAT
Journal Home
Volume Issue
Issue
Volume-3, Issue-3, Spl. Iss-2  ( Sep, 2015 )
Paper Title
A Method For Exemplar Based Inpainting By Combining Graph Based Segmentation And Expectation Maximization Algorithm
Author Name
Ajana Ajayakumar, Monisha Menon
Affilition
Department of Electronics &Communication
Pages
70-75
Abstract
Image inpainting is a process of restoring lost, damaged or selected portion of an image based on the neighboring information. The inpainted result should be such that, when viewed by any ordinary observer should feel a visually pleasing flow of data in and around the hole (selected region). Image Inpainting algorithm have a number of applications such as rebuilding of damaged photographs & films, removal of stamped date from photographs, removal of unwanted objects, red eye correction etc. This paper proposes a method for image inpainting using DEMA (Diffused Expectation Maximization Algorithm) and GBS (Graph Based Segmentation). Here an Exemplar-based in-painting algorithm is adopted, which iteratively search the neighboring region (source region) and fill the missing region (target region), with the most suited patches from the source region. Here the proposed method uses segmentation map along with diffused expectation maximization algorithm to improve the performance of inpainting. In Graph Based Segmentation method the spatial information in the source region is utilized .This method selects the parameter values of the robust priority function and thereby determines the most suited patch size and reduces the search region. The number of segments obtained from GBS is used in DEMA thereby we get a DEMA segmented image which is once again graph based segmented in order to get the parameter value of the DEMA segmented image. This parameter value along with the segmented image is passed through an exemplar based inpainting algorithm forming a better inpainted image. Certain parameters to evaluate the performance of the proposed algorithm is also done. Experimental results with a number of test images shows the effectiveness of the proposed method. Index Terms- Diffused Expectation Maximization, Exemplar Based Inpainting, Garph Based Segmentation, Inpainting.
  View Paper