Pages

Showing posts with label Papers. Show all posts
Showing posts with label Papers. Show all posts

Saturday, January 31, 2009

Fast Image Retrieval Based on Equal-average Equal-variance K-Nearest Neighbour Search

This paper presents two fast schemes to speed up the retrieval process for conventional content-based image retrieval systems. The traditional features such as color and invariant histograms are extracted offline from each image to compose a feature vector. All these feature vectors construct the feature database. Then the system performs the online retrieval based on this database as soon as possible. In the case of a small number of returned images, an equal-average equalvariance K nearest neighbour search (EEKNNS) method is used to speed up the retrieval process. In the case of a large number of returned images, an iterative EEKNNS (IEEKNNS) method is given. Experimental results show that the proposed retrieval methods can largely accelerate the retrieval process while guaranteeing the same recall and precision.

Read More

Wednesday, January 7, 2009

Private Content Based Image Retrieval

For content level access, very often database needs the query as a sample image. However, the image may contain private information and hence the user does not wish to reveal the image to the database. Private Content Based Image Retrieval (PCBIR) deals with retrieving similar images from an image database without revealing the content of the query image – not even to the database server. We propose algorithms for PCBIR, when the database is indexed using hierarchical index structure or hash based indexing scheme. Experiments are conducted on real datasets with popular features and state of the art data structures. It is observed that specialty and subjectivity of image retrieval (unlike SQL queries to a relational database) enables in computationally efficient yet private solutions.
Author: Shashank J, Kowshik P, Kannan Srinathan and C.V. Jawahar
Download Tha Paper (TR)

CONTENT BASED ONLINE LOOKUP OF DATA SETS WITH PCA

This paper presents an approach to unsupervised, content based classification and lookup of multi-dimensional objects such as pictures, audio and video, which could be used to accelerate online searches of such data sets and real-time biometrical recognition and classification which uses them. The approach treats the classification and lookup as a generic pattern matching problem. On contrary to the traditional correlation oriented method, this approach adopts Principle Component Analysis (PCA) to create a representation which will be used as an index to a particular data set. This results in a fewer dimension identity of the object, that can be looked up in shorter time, provided the approach always consider the first n components for matching. Further the outcome of PCA will be classified through dendrogram clusterization that will result in a system similar to a hash table. The framework provides hash function like system to speedup lookups and recognitions. The article introduces to the problem of classification of non-textual data and inherent limitations. It further describes the advantage of using PCA as an alternative to pattern matching and leads through the methodology in developing dendrogram clusterization based online management and lookup.
Autor: Gartheeban Ganeshapillai
Download the paper