Incremental codebook generation for vector quantization in large scale content based image retrieval

2010 ◽  
Author(s):  
B. Janet ◽  
A. V. Reddy ◽  
S. Domnic
Keyword(s):  
Image Retrieval ◽  
Vector Quantization ◽  
Large Scale ◽  
Content Based Image Retrieval ◽  
Codebook Generation

scholarly journals Fully Automated Pose Estimation of Historical Images in the Context of 4D Geographic Information Systems Utilizing Machine Learning Methods

2021 ◽  
Vol 10 (11) ◽  
pp. 748
Author(s):  
Ferdinand Maiwald ◽  
Christoph Lehmann ◽  
Taras Lazariv
Keyword(s):  
Information Systems ◽  
Geographic Information Systems ◽  
Image Retrieval ◽  
Pose Estimation ◽  
Large Scale ◽  
Feature Matching ◽  
Geographic Information ◽  
Content Based Image Retrieval ◽  
Matching Methods ◽  
Historical Images

The idea of virtual time machines in digital environments like hand-held virtual reality or four-dimensional (4D) geographic information systems requires an accurate positioning and orientation of urban historical images. The browsing of large repositories to retrieve historical images and their subsequent precise pose estimation is still a manual and time-consuming process in the field of Cultural Heritage. This contribution presents an end-to-end pipeline from finding relevant images with utilization of content-based image retrieval to photogrammetric pose estimation of large historical terrestrial image datasets. Image retrieval as well as pose estimation are challenging tasks and are subjects of current research. Thereby, research has a strong focus on contemporary images but the methods are not considered for a use on historical image material. The first part of the pipeline comprises the precise selection of many relevant historical images based on a few example images (so called query images) by using content-based image retrieval. Therefore, two different retrieval approaches based on convolutional neural networks (CNN) are tested, evaluated, and compared with conventional metadata search in repositories. Results show that image retrieval approaches outperform the metadata search and are a valuable strategy for finding images of interest. The second part of the pipeline uses techniques of photogrammetry to derive the camera position and orientation of the historical images identified by the image retrieval. Multiple feature matching methods are used on four different datasets, the scene is reconstructed in the Structure-from-Motion software COLMAP, and all experiments are evaluated on a newly generated historical benchmark dataset. A large number of oriented images, as well as low error measures for most of the datasets, show that the workflow can be successfully applied. Finally, the combination of a CNN-based image retrieval and the feature matching methods SuperGlue and DISK show very promising results to realize a fully automated workflow. Such an automated workflow of selection and pose estimation of historical terrestrial images enables the creation of large-scale 4D models.

scholarly journals Towards Large Scale Image Retrieval System Using Parallel Frameworks

2020 ◽  
Author(s):  
Saliha Mezzoudj
Keyword(s):  
Big Data ◽  
Image Retrieval ◽  
Large Scale ◽  
Retrieval System ◽  
Image Databases ◽  
Content Based Image Retrieval ◽  
Content Search ◽  
Parallel Frameworks ◽  
Image Retrieval System ◽  
Large Scale Image Retrieval

Recently, the increasing use of mobile devices, such as cameras and smartphones, has resulted in a dramatic increase in the amount of images collected every day. Therefore, retrieving and managing these large volumes of images has become a major challenge in the field of computer vision. One of the solutions for efficiently managing image databases is an Image Content Search (CBIR) system. For this, we introduce in this chapter some fundamental theories of content-based image retrieval for large scale databases using Parallel frameworks. Section 2 and Section 3 presents the basic methods of content-based image retrieval. Then, as the emphasis of this chapter, we introduce in Section 1.2 A content-based image retrieval system for large-scale images databases. After that, we briefly address Big Data, Big Data processing platforms for large scale image retrieval. In Sections 5, 6, 7, and 8. Finally, we draw a conclusion in Section 9.

A Multi-resolution Approach for Content-based Image Retrieval on the Grid

2005 ◽  
Vol 44 (02) ◽  
pp. 211-214 ◽  
Author(s):  
T. Tweed ◽  
S. Miguet ◽  
K. Hassan
Keyword(s):  
Image Retrieval ◽  
High Performance ◽  
Large Scale ◽  
Breast Cancer Diagnosis ◽  
Content Based Image Retrieval ◽  
Distributed Data ◽  
Promising Tool ◽  
Medical Centers ◽  
Image Retrieval System ◽  
Processing Resources

Summary Objectives: Hospitals and medical centers are producing more and more data that need to be processed. Those data are confidential, heterogeneous, and limited to the geographic site where they have been produced. Unless properly anonymized, they cannot be distributed on wide area networks. Methods: Grid technologies allow the globalization of storage and processing resources, and enable large-scale experimentations on distributed data. They constitute a promising tool to treat the different data and analyze the knowledge they contain, while offering secured access and high-performance computing capacities to the different users. Our aim is to evaluate the possibilities of grid technologies for handling medical data. Results and Conclusions: In this paper, we focus on a breast cancer diagnosis assistance tool, based on distributed and incremental knowledge construction and a content-based image retrieval system. We analyze the different scenarios of uses of such a tool. We further propose an algorithm that indexes mammographic images for content-based query purposes. This algorithm is tested on images of different resolutions in order to reduce the indexation time and we analyze its performance with experiments on the grid.

scholarly journals CONTENT-BASED IMAGE RETRIEVAL UNTUK PENCARIAN PRODUK PONSEL

2019 ◽  
Vol 3 (1) ◽  
pp. 31
Author(s):  
Nickolas Cornelius Siantar ◽  
Jaqnson Hendryli ◽  
Dyah Erny Herwindiati
Keyword(s):  
Image Retrieval ◽  
Vector Quantization ◽  
Euclidean Distance ◽  
Local Binary Pattern ◽  
Feature Vector ◽  
Content Based Image Retrieval ◽  
Color Quantization ◽  
Histogram Of Oriented Gradient ◽  
Feature Extractor ◽  
Online Stores

Phone or smartphone and online shop, there is something that cannot be separated with human. There are so many type of smartphones show up in the market that people are confused on which one to get on the online stores. Smartphones recognition is done by using the Histogram of Oriented Gradient to recognize shapes of phones, Color Quantization to recognize the color, and Local Binary Pattern to recognize texture of the phones. The output of the Feature Extractor is a feature vector which is used on the LVQ to process recognize through finding the smallest Euclidean Distance between the trained vectors. The result of this paper is an application that can recognize 16 phone types using the image with the accuracy of 9.6%. Pada saat ini, ponsel dan toko online merupakan sesuatu yang tidak dapat dipisahkan dari manusia. Begitu banyak jenis ponsel bermunculan setiap tahunnya sehingga menyebabkan manusia bingung dalam mengenali ponsel tersebut. Pada program pengenalan ponsel ini digunakan Histogram of Oriented Gradient untuk mengambil fitur berupa bentuk ponsel, Color Quantization untuk mengambil fitur warna, dan Local Binary Pattern untuk mengambil fitur tekstur ponsel. Hasil dari pengambilan fitur berupa fitur vektor yang digunakan pada Learning Vector Quantization untuk proses pengenalan dengan mencari nilai terkecil Euclidean Distance antara vektor fitur dengan vektor bobot terlatih. Hasil dari program pengenalan ini yaitu program dapat melakukan pengenalan terhadap 16 jenis ponsel dengan akurasi sebesar 9.6%.

scholarly journals Multidimensional Markov Stationary Feature for Image Retrival Systems

2016 ◽  
Vol 44 ◽  
pp. 113-122
Author(s):  
Md. Saiful Islam ◽  
Md. Emdadul Haque ◽  
Md. Ekramul Hamid
Keyword(s):  
Image Retrieval ◽  
Large Scale ◽  
Experimental Result ◽  
Content Based Image Retrieval ◽  
Retrieval Algorithm ◽  
Histogram Method

Markov Stationary Features (MSF) not only considers the distribution of colors like histogram method does, also characterizes the spatial co-occurrence of histogram patterns. However, handling large scale database of images, simple MSF method is not sufficient to discriminate the images. In this paper, we have proposed a robust content based image retrieval algorithm that enhances the discriminating capability of the original MSF. The proposed Multidimensional MSF (MMSF) algorithm extends the MSF by generating multiple co-occurrence matrices with different quantization levels of an image. Publicly available WANG1000 and Corel10800 databases are used to evaluate the performance of the proposed algorithm. The experimental result justifies the effectiveness of the proposed method.

Sign in / Sign up

Export Citation Format

Share Document