China
Africa
Explore chapters and articles related to this topic
The KD-ORS Tree: An Efficient Indexing Technique for Content-Based Image Retrieval
Published in D. P. Acharjya, V. Santhi, Bio-Inspired Computing for Image and Video Processing, 2018
The amount of multimedia data has strongly increased in recent years. Therefore, new efficient and powerful applications are needed to handle the data. Particularly, most applications require efficient methods to retrieve the relevant data. In such situations, indexing of high dimensions is important. The idea of indexing is mapping the extracted descriptor from images into a high-dimensional space. Each image in the image databases is described using visual features such as color, texture, and shape. So, image signatures have high dimensions to represent the images. To find the similar images from the image database, an efficient similarity search plays a significant role. In order to achieve an efficient similarity search in an image database, a robust method to index the high-dimensional feature space is to be developed. Similarity search corresponds to range search on the indexed structure. The distance between two vectors is frequently used to estimate the similarity between the related images. Therefore, the problem of finding the most similar images to a given query image can be seen as a problem of k-nearest neighbour (k-NN) search in high-dimensional vector space. The methods that have been proposed for searching are known as high-dimensional indexing methods.
Content-Based Image Retrieval Techniques
Published in Wahiba Ben Abdessalem Karaa, Nilanjan Dey, Mining Multimedia Documents, 2017
Sayan Chakraborty, Prasenjit Kumar Patra, Nilanjan Dey, Amira S. Ashour
In a multimedia retrieval system, similarity search plays a vital role. This technique is widely used in scientific as well as commercial applications such as near-duplicate detection of images and videos or CBIR-based audio, video, or image retrieval. The main challenges of data objects’ inherent properties are collected using feature representation. It has been observed that any similarity measurement [22] framework can define the similarity between the query and the target object from the database. This is done by measuring the distance between the corresponding feature representations. These distance values can be further processed to retrieve mostly similar objects from the database.
Multimedia Storage
Published in Sreeparna Banerjee, Elements of Multimedia, 2019
Similarity search [9] is the process of locating a record in a database that is closest to the query. Similarity searching is performed using nearest neighbor finding. Existing methods fall into two classes: A mapping to a low dimension vector space, a process known as dimensionality reduction, is performed. This is then indexed using representations such as quad trees, kD trees, and R-trees.Mapping and indexing of objects is based on distances using representations such as VP tree and Mtree.
BTLA-LSDG: Blockchain-Based Triune Layered Architecture for Authenticated Subgraph Query Search in Large-Scale Dynamic Graphs
Published in IETE Journal of Research, 2023
G. Sharmila, M. K. Kavitha Devi
Chae et al. [23] propose a new concept called graph bag, which is comprised of web pages, XML documents, images and drugs. Dynamic bag classification is implemented in this paper and a very informative feature set (top feature set) is considered to deal with the stream of bags. Finally it classifies the dynamic bags. Feature selection at each time point increases the time complexity. In [24] a similarity search is implemented for large graph databases by a fast algorithm. The problem of K-NN (K-Nearest Neighbor)-based similarity search is concentrated and Locality Sensitive Hashing (LSH) is proposed. For each query, Euclidean distance space is invoked with the graph database. It is conducted for image retrieval and images are considered to be attributed to the graph. There are two challenges that are unfilled in this paper, namely, vector-based graph representation consumes larger time and the LSH method requires high storage space.