A Mighty Image Retrieval Descriptor Based on Machine Learning and Gaussian Derivative Filter

Aug 4, 2026

Upload by Oliver

A machine learning–enhanced image descriptor is presented to improve content-based image classification and retrieval performance. The approach integrates a Gaussian derivative filter scaffold (GDF-HOG) with an enhanced AlexNet convolutional neural network and applies principal component analysis to reduce feature dimensionality. Experiments are conducted on Oliva and Torralba, Caltech-101, Wang, and Coil100 datasets, reporting strong recognition accuracy and demonstrating average improvements versus other descriptor-based classifiers. Results also indicate higher precision with reduced computational complexity.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages9

Reading Time23 min

AccessFree to download

Introduction

Related Work

Proposed System

Experimental Results

Conclusion and Future Perspectives

A Mighty Image Retrieval Descriptor Based on Machine Learning and Gaussian Derivative Filter Original Scientific Paper El Aroussi El Mehdi Chouaib Doukkali University, ELITES Laboratory, Departement of Computer Science and Mathematics Higher School of Technology El Jadida, Morocco [Elaroussi.e@ucd.ac.ma](Elaroussi.e@ucd.ac.ma) Barakat Latifa Chouaib Doukkali University, Management of Sustainable Agriculture Laboratory, Higher School of Technology El Jadida, Morocco [barakatlati@gmail.com](barakatlati@gmail.com) Silkan Hassan Chouaib Doukkali University, LaROSERI Laboratory, Department of computer sciences, Faculty of Sciences, El Jadida, Morocco [silkan_h@yahoo.fr](silkan_h@yahoo.fr) Abstract – The development of new image descriptor has always been an important topic to improve the efficiency of contentbased image classification and retrieval. Improvements and developments in machine learning and deep learning algorithms as well as artificial intelligence algorithms are widely used by researchers to obtain effective CBIR descriptors. In our article, we will present a robust image descriptor, extended by machine learning and deep learning algorithms. The descriptor is provided through a Gaussian derivative filter scaffold named GDF-HOG with an enhanced convolutional neural network (CNN) AlexNet, to reduce the dimensions we used the principal component analysis algorithm. The experimental results were carried out on Oliva and Torralba, Caltech-101, Wang and Coil100 datasets. Experiments show that the accuracy of the proposed method is 98.23% for Coil-100%, 95.92% for Corel-1000, value 87.17 and 94.6% for Oliva and Torralba. In comparison our results with other descriptor imageclassifiers show that they achieved accuracy increases of 0.12% on average and up to 3.23%. These experimental results affirm the advantage of the proposed descriptor over existing systems based in terms of average accuracy. the proposed descriptor improves the precision, and also reduces the complexity of the calculation. Keywords: Federated Learning, Machine Learning, Deep Learning, Privacy, Collaborative Machine Learning Received: Received: August 12, 2023; Received in revised form: October 19, 2023; Accepted: October 20, 2023 1. INTRODUCTION Nowadays, the amount of digital images in the form of personalized and corporate collections has increased enormously thanks to the widespread and easy use of the Internet and the enormous use of audiovisual data in digital format for communications. Hence, there is a increasing demand for powerful image indexing and retrieval in an automatic way. Nevertheless, with such use and availability of images, the solutions based on textual images (grace of keywords) become impassable and inappropriate for indexing and retrieving images. To overcome this problem content-based image retrieval (CBIR) has become a great research interest among re- search communities [1,2,3]. Content-based image indexing and search discriptors generate considerable image representations by considering the visual features of images, i.e. salient points ,texture and shape [4-14], It brings similar images using distance as a semantic result. The audacious increase in image descriptors has been an active field of research and will help to increase the performance vast actions in computer vision. various systems such as cale-invariant feature transform (SIFT) [15], speeded up robust features (SURF) [16],cooccurrence matrix (GLCM) [45] , local binary patterns (LBP) [17], GIST algorithms were used for CBIR systems [18]. Such hand-crafted feature generation algorithms are still used in Machine Learning [19–22]. Each of these Volume 15, Number 5, 2024 427 descriptors has abnormalities, such as a large capacity of the feature vector , is that it cannot describe the characteristics of textures efﬁciently and distinctively and is mathematically weak and sensitive to noise. In this article we propose an efficient i

FAQs

What is the main goal of the proposed method?

To build a robust image descriptor that improves content-based image indexing and retrieval performance using machine learning and deep learning components.

How does the descriptor combine GDF-HOG and AlexNet?

The descriptor uses a Gaussian derivative filter scaffold (GDF-HOG) together with an enhanced CNN based on AlexNet, then applies PCA for dimensionality reduction.

Which datasets are used to evaluate the method and what results are reported?

Evaluation is performed on datasets including Oliva and Torralba, Caltech-101, Wang, and Coil100, with reported accuracies reaching 98.23% on Coil-100 and showing up to 3.23% average accuracy improvement versus existing descriptor-based classifiers.

Aug 4, 2026

Upload by Oliver