CONTENT BASED IMAGE RETRIEVAL (CBIR) SYSTEM

Advancement in hardware and telecommunication technology has boosted up creation
and distribution of digital visual content. However this rapid growth of visual content
creations has not been matched by the simultaneous emergence of technologies to support
efficient image analysis and retrieval. Although there are attempt to solve this problem by
using meta-data text annotation but this approach are not practical when it come to the
large number of data collection.
This system used 7 different feature vectors that are focusing on 3 main low level feature
groups (color, shape and texture). This system will use the image that the user feed and
search the similar images in the database that had similar feature by considering the
threshold value. One of the most important aspects in CBIR is to determine the correct
threshold value. Setting the correct threshold value is important in CBIR because setting
it too low will result in less image being retrieve that might exclude relevant data. Setting
to high threshold value might result in irrelevant data to be retrieved and increase the
search time for image retrieval.
Result show that this project able to increase the image accuracy to average 70% by
combining 7 different feature vector at correct threshold value.
iii