Low bitrate image retrieval with compressed histogram of gradients descriptors
Abstract/Contents
- Abstract
- Mobile Visual Search (MVS) enables a new class of applications which use the camera on mobile devices like smart phones and tablets to initiate search queries. For MVS applications, transmitting a JPEG image could take tens of seconds over a slow wireless link. The size of the data transmitted over the network needs to be as small as possible to reduce network latency. In this thesis, we show that we can reduce the amount of data by an order of magnitude by transmitting compressed features compared to transmitting JPEG images. Most popular descriptors in computer vision are based on histograms (or distributions) of gradients.To that end, we study techniques for quantization and coding of distributions, and use these techniques to obtain compact representations for HoG descriptors. The proposed descriptors are fast to compute, do not require storage of codebooks and can be compared in the compressed domain. We discuss how to build an analytical model for each step of an image matching and retrieval pipeline. We show how image level True Positive Rate and False Positive Rate can be modeled from feature level Receiver Operating Characteristic. We also propose a model for the bag-of-words retrieval pipeline. We use the model for optimizing system parameters in a low bitrate image retrieval system. We show that high performance retrieval can be achieved with a database of 1 million images using only a few KB of compressed feature query data.
Description
Type of resource | text |
---|---|
Form | electronic; electronic resource; remote |
Extent | 1 online resource. |
Publication date | 2013 |
Issuance | monographic |
Language | English |
Creators/Contributors
Associated with | Chandrasekhar, Vijay Ramaseshan |
---|---|
Associated with | Stanford University, Department of Electrical Engineering. |
Primary advisor | Girod, Bernd |
Thesis advisor | Girod, Bernd |
Thesis advisor | Grzeszczuk, Radek, 1967- |
Thesis advisor | Guibas, Leonidas J |
Advisor | Grzeszczuk, Radek, 1967- |
Advisor | Guibas, Leonidas J |
Subjects
Genre | Theses |
---|
Bibliographic information
Statement of responsibility | Vijay Chandrasekhar. |
---|---|
Note | Submitted to the Department of Electrical Engineering. |
Thesis | Thesis (Ph.D.)--Stanford University, 2013. |
Location | electronic resource |
Access conditions
- Copyright
- © 2013 by Vijay Ramaseshan Chandrasekhar
- License
- This work is licensed under a Creative Commons Attribution Non Commercial 3.0 Unported license (CC BY-NC).
Also listed in
Loading usage metrics...