Private Content Based Image Retrieval

1
Private Content Based Image Retrieval
Shashank J, Kowshik P, Kannan Srinathan and C.V.
Jawahar
Is it possible for an image database to respond
accurately without any knowledge of the query
image.

• Objective
• Retrieve results from an image database, while
  maintaining complete privacy of the query image
  from the database.
• Applications
• Medical Image Databases
• Surveillance Systems
• Logo Patent Search
• Defense Systems
• Web 2.0
• Image based query retrieval

Extension to other Hierarchical Structures
Results and Discussions
Quadratic Residuosity Assumption

• KD Tree and Corel Dataset
• Color Histogram (768 dimensions)
• Retrieval Time 0.596 secs
• Vocabulary Tree and Nister Dataset
• SIFT features
• Vocabulary of 10000 visual words.
• Retrieval Time 0.320 secs

• Hierarchical Structures vary in
• Number of nodes at each level.
• Information at a node.
• Any number of nodes can be converted into a m x
  n matrix.
• Any information can be represented in binary
  format.
• If the user has the data about the indexing
  structure and the format of the information
  stored at a node, the algorithm can be simulated
  for any hierarchical structure.
• KD Tree
• Similar to binary tree
• Each node contains split dimension and split
  value.
• Vocabulary Tree
• Branch factor depends on vocabulary size.

Quadratic Residuosity Assumption Given a
number y YN, it is predictably hard to decide
whether y is a QR or a QNR. Basic
Properties QNR x QNR QR QNR x QR QNR QR
x QR QR
PCBIR Algorithm

1. Extract feature vector of the query image say
   fquery.
2. The user first asks the database to send the
   information at the root node.
3. Using fquery and the information received, the
   user decides whether to access the left subtree
   or the right subtree.
4. In order to get the data at the node to be
   accessed, the user frames a query Qi where i
   indicates the level in which the node occurs.
5. The database returns a reply Ai for the query Qi.
6. The user using the data obtained from Ai and
   fquery decides which subtree to move next to.

PCBIR on a Binary Search Tree
Formulation of Qi and Ai
Center for Visual Information Technology Internati
onal Institute of Information Technology,
Hyderabad, INDIA http//cvit.iiit.ac.in