Authors :
K. B. Ranushka Pasindu Dharmaranga; Ligitha Sakthymayuran
Volume/Issue :
Volume 7 - 2022, Issue 5 - May
Google Scholar :
https://bit.ly/3IIfn9N
Scribd :
https://bit.ly/3NBzadh
DOI :
https://doi.org/10.5281/zenodo.6787485
Abstract :
Image reconstruction is currently often used
in a wide range of technological and medical
applications. The local image feature descriptor is the
most critical factor influencing the performance of
object reconstruction or image retrieval systems. This
study provides and demonstrates a strategy for
replicating images. In this approach, Training photos are
used to extract local feature descriptors; at first images
are recreated using local feature descriptors and
geometric information. Scale invariant feature transform
(SIFT) descriptors are used to characterize images, and
the feature extraction method is similar to how
descriptors are used in the training phase. The unknown
image closest neighbor descriptor built by using pairwise
matching. For each of the regions of interest, visually
equivalent patches may be in the external image
database.To detect patch overlapping regions between
the new patch and the patch already present in the query
image, the Mean Squared Error (MSE) is used. To
eliminate overlapping patches, the highest MSE
threshold value is chosen as the default threshold (DT) in
this experimental technique. Based on the experimental
results, an image may be approximated and rebuilt using
image local feature descriptors.
Keywords :
Image reconstruction, image retrieval, image feature descriptor, geometric information, partial information.
Image reconstruction is currently often used
in a wide range of technological and medical
applications. The local image feature descriptor is the
most critical factor influencing the performance of
object reconstruction or image retrieval systems. This
study provides and demonstrates a strategy for
replicating images. In this approach, Training photos are
used to extract local feature descriptors; at first images
are recreated using local feature descriptors and
geometric information. Scale invariant feature transform
(SIFT) descriptors are used to characterize images, and
the feature extraction method is similar to how
descriptors are used in the training phase. The unknown
image closest neighbor descriptor built by using pairwise
matching. For each of the regions of interest, visually
equivalent patches may be in the external image
database.To detect patch overlapping regions between
the new patch and the patch already present in the query
image, the Mean Squared Error (MSE) is used. To
eliminate overlapping patches, the highest MSE
threshold value is chosen as the default threshold (DT) in
this experimental technique. Based on the experimental
results, an image may be approximated and rebuilt using
image local feature descriptors.
Keywords :
Image reconstruction, image retrieval, image feature descriptor, geometric information, partial information.
