Content-based image retrieval based on combination of texture and colour information extracted in spatial and frequency domains
| Pages | 650-666 |
| Date | 05 August 2019 |
| Published date | 05 August 2019 |
| DOI | https://doi.org/10.1108/EL-03-2019-0067 |
| Author | Neda Tadi Bani,Shervan Fekri-Ershad |
| Subject Matter | Information & knowledge management,Information & communications technology,Internet |
Content-based image retrieval
based on combination of texture
and colour information extracted
in spatial and frequency domains
Neda Tadi Bani
Faculty of Computer Engineering, Najafabad Branch, Islamic Azad University,
Najafabad, Iran, and
Shervan Fekri-Ershad
Faculty of Computer Engineering, Najafabad Branch, Islamic Azad University,
Najafabad, Iran and Big Data Research Center, Najafabad Branch,
Islamic Azad University, Najafabad, Iran
Abstract
Purpose –Large amount of data are stored in image format. Image retrieval from bulk databases has
become a hot research topic. An alternative method for efficient image retrieval is proposed based on a
combination of texture and colour information. The main purpose of this paper is to propose a new content
based image retrieval approachusing combination of color and texture information in spatial and transform
domainsjointly.
Design/methodology/approach –Various methods are provided for image retrieval, which try to
extract the image contentsbased on texture, colour and shape. The proposed image retrievalmethod extracts
global and local texture and colour informationin two spatial and frequency domains. In this way, image is
filtered by Gaussian filter, then co-occurrence matrices are made in different directions and the statistical
features are extracted. The purpose of this phase is to extract noise-resistant local textures. Then the
quantised histogramis produced to extract global colour information in the spatial domain.Also, Gabor filter
banks are used to extract local texture featuresin the frequency domain. After concatenating the extracted
featuresand using the normalised Euclidean criterion, retrieval is performed.
Findings –The performance of the proposed method is evaluated based on the precision, recall and run
time measures on the Simplicity database. It is compared with many efficient methods of this field. The
comparisonresults showed that the proposed methodprovides higher precision than many existingmethods.
Originality/value –The comparison resultsshowed that the proposed method provides higher precision
than many existing methods. Rotation invariant, scale invariant and low sensitivity to noise are some
advantages of the proposed method.The run time of the proposed method is within the usual time frame of
algorithmsin this domain, which indicates that the proposedmethod can be used online.
Keywords Content-based image retrieval, Texture analysis, Colour histograms, Statistical features,
Co-occurrence matrices, Gabor filters
Paper type Research paper
1. Introduction
The need to find relevant images to a query image from huge databases is an active topic
research in computer vision, whichis called image retrieval (Tyagi,2018). The need to find a
desired image from a huge collection is shared by many professional groups, such as
journalists, forensic experts, design engineers, art historiographers, and so on. Traditional
EL
37,4
650
Received14 March 2019
Revised19 May 2019
Accepted4 July 2019
TheElectronic Library
Vol.37 No. 4, 2019
pp. 650-666
© Emerald Publishing Limited
0264-0473
DOI 10.1108/EL-03-2019-0067
The current issue and full text archive of this journal is available on Emerald Insight at:
www.emeraldinsight.com/0264-0473.htm
image retrieval engines index multimedia visual data using the surrounding metadata
information arounddatabase or web images, such as titles, tags, texts, or keywords (Li et al.,
2016). These engines are usually called concept-based image retrieval or text-based retrieval
engines. Content-based image retrieval (CBIR) is opposed to traditional concept-based
approaches. Content-basedmeans that the search engine analyses the contents of the image
rather than the metadata. The term content in this description refers to textures, colours,
shapes, or any other visual information that can be derived from the image. Texture or
colour information of the query image may be inconsistent with the visual real content,
CBIR is preferred andhas been witnessed to make great advances in recent years.
A wide range of possible applications for CBIR technology has been identified.
Potentially fruitfulareas include:
crime prevention;
intellectual property;
architectural and engineering design;
fashion and interior design;
journalism and advertising;
medical diagnosis;
geographical information and remote sensing systems;
cultural heritage;
education and training;
home entertainment; and
Web searching.
There are various methods for image retrieval. Most of them try to extract the texture and
colour information of the image simultaneously to increase the system’sfinal accuracy. A
brief increase in system precisionprovides high economic benefits to search engine owners.
The point that some researchersdo not pay attention to is the extraction of local and general
features of the image simultaneously. Research results in Ismail (2017) showed that the
analysis of image components, such as texture and colour, in both spatial and frequency
domains, can increasethe accuracy of the image classification method.
In this regard, this paper presents a way to retrieve images based on the combination of
texture and colour information. Contrary to some of the available CBIR techniques, texture and
colour features, both locally and generally, are extracted. Also, extraction features have the
ability to define the content of the image; information is used in both the spatial and frequency
domains. The feature extraction phase in this paper consists of three main sub-steps. In the first
sub-step, to extract the local noise-resistant texture information, the input image is first filtered
by a Gaussian filter. Then the co-occurrence matrixes are calculated and constructed in different
directions. Then the statistical features of the matrices are extracted. The purpose of the second
sub-step is to extract the general colour information in the spatial domain. For this purpose, a
colour-coded histogram is used. The purpose of the third sub-step is to extract texture
informationinthefrequencydomain.For this purpose, the Gabor filter bank has been used.
Typically, methods that have been presented so far for image retrieval only analyse
image information locallyor globally. There are fewer articles that extract these two groups
of features together. Many researchers believe that the combination of local and general
information in two domains (spatial and transform)increases the computational complexity.
Extracting a discriminative sub-set of texture and colour information in spatial and
Combination
of texture and
colour
information
651
To continue reading
Request your trial