Images retrieval/deletion using VP-Tree Index Structure
Introduction:
An image database is built using VP-Tree index structure. There are 1490
images in the database. User can retrieve images in 3 different ways. One
is input a range. The general idea is the larger the value, more but less
similar pictures will be found. The smaller the value, fewer but more
similar pictures will be found. As the number of pictures found cannot be
expected, another method is provided. This method is to input the number of
pictures to be found. The program will search for the number of pictures
which are most similar to the query. The above two methods always retrieve
similar images, so the third method is provided to explore different images.
This method can retrieve non-duplicate images from the database randomly.
Through this method, user can explore completely different images. With the
combination uses of above method, a desired images is selected and it can be
deleted from the database.
There are two sets of features for the images. The first set is 27
dimensions. The other set is 9 dimensions which is obtained by the
KL-transformation of the first set of features. Though this demonstration,
user can see that the transformed features preserved most of the information
of the original one.
Enter the number of images to get randomly
thip@cs.cuhk.hk & chmak2@cs.cuhk.hk