Content-Based Image Retrieval through Improved Subblock Technique
Traditional Content-Based Image Retrieval (CBIR) systems mainly relied on the extraction of features globally. The drawback of this approach is that it cannot sufficiently capture the important features of individual regions in an image which users might be interested in. Due to that, an extensio...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English English |
| Published: |
2006
|
| Subjects: | |
| Online Access: | http://psasir.upm.edu.my/id/eprint/5196/1/FSKTM_2006_6.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Traditional Content-Based Image Retrieval (CBIR) systems mainly relied on the
extraction of features globally. The drawback of this approach is that it cannot
sufficiently capture the important features of individual regions in an image which
users might be interested in. Due to that, an extension of the CBIR systems is
designed to exploit images at region or object level. One of the important tasks in
CBIR at region or object level is to segment images into regions based on low-level
features. Among the low-level features, colour and location information are widely
used.
In order to extract the colour information, Colour-based Dominant Region
segmentation is used to extract a maximum of three dominant colour regions in an
image together with its respective coordinates of the Minimum-Bounding Rectangle (MBR). The Sub-Block technique is then used to determine the location of the
dominant regions by comparing the coordinates of the region’s MBR with the four
corners of the centre of the location map. The cell number that is maximally covered
by the region is supposedly to be assigned as the location index. However, the Sub-
Block technique is not reliable because in most cases, the location index assigned is
not the cell number that is maximally covered by the region and sometimes a region
does not overlap with the cell number assigned at all. The effectiveness of this
technique has been improved by taking into consideration the total horizontal and
vertical distance of a region at each location where it overlaps. The horizontal
distance from the left edge to the right edge of a region and the vertical distance
from the top edge to the bottom edge of a region are calculated. The horizontal and
vertical distances obtained for that region are then counted. The cell number with the
highest distance would be assigned as the location index for that region. The
individual colour and location index of each dominant region in an image is
extended to provide combined colour-spatial indexes. During retrieval, images in the
image database that have the same index as the query image is retrieved.
A CBIR system implementing the Improved Sub-Block technique is developed. The
CBIR system supports Query-By-Example (QBE). The retrieval effectiveness of the
improved technique is tested through retrieval experiments on six image categories
of about 900 images. The precision and recall is measured. From the experiments it
is shown that retrieval effectiveness has been significantly improved by 85.86%
through the Improved Sub-Block technique. |
|---|