As the volume of digital data continues to grow exponentially, DNA has emerged as a promising medium for long-term data storage due to its high density and durability. For enabling data retrieval via DNA's biochemical reactions, the encoding strategy plays a critical role. This paper proposes a training framework for a DNA encoder that improves both accuracy and training efficiency in content-based image retrieval by incorporating deep metric learning. In addition, we introduce loss functions that enforce biological constraints, specifically homopolymer length and GC content, thereby improving the biochemical stability of the generated DNA sequences. To evaluate the effectiveness of the proposed method, we conduct quantitative assessments based on image classification performance. Simulations on the CIFAR- 10 and CIFAR-100 datasets demonstrate that our method achieves classification accuracy comparable to CNN-based baselines and a 20- fold speedup over the training time of the existing method. Moreover, the generated DNA sequences enable strict control of homopolymer length and maintain GC content within the optimal 40-60 improving biological feasibility compared to baseline methods. The source code is publicly available at GitHub.

As digital imaging in healthcare grows quickly, dealing with vast medical image data is getting trickier. Content-Based Medical Image Retrieval (CBMIR) systems help with this, but they struggle because of the gap between simple image details and what these images mean in a clinical setting. This paper presents a new approach using deep learning for CBMIR that combines Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs), and Explainable AI (XAI). Using the Breast Ultrasound Image (BUSI) dataset for training, this hybrid model classifies images and finds the relevant results based on predictions. It reaches a classification accuracy of 99.24% and performs well in retrieval tasks.

Texture image retrieval based on subjective visual descriptions remains a significant challenge due to the “semantic gap”, where conventional Content-Based Image Retrieval (CBIR) methods rely on low-level features or reference images that often diverge from human perception. To bridge this gap, this paper proposes a reference-free, perception-driven retrieval framework that enables users to query textures directly via abstract perceptual attributes. First, we constructed a human-centric perceptual feature space through controlled psychophysical experiments, quantifying 12 explicit texture attributes (e.g., granularity, directionality) using a 9-point Likert scale. Second, addressing the variability in visual sensitivity across user demographics, we developed a user-adaptive mechanism incorporating dual perceptual libraries tailored for art-major and non-art-major groups. Retrieval is formulated as a perception-aligned similarity optimization problem within this normalized space. Experimental evaluations on the Describable Textures Dataset (DTD) demonstrate that our method achieves superior perceptual consistency compared to both handcrafted descriptors (GLCM, LBP, HOG) and deep learning baselines (VGG16, ResNet50). Notably, the framework attained high PAP@3 performance across both user groups, validating its effectiveness in decoding fuzzy human intent without the need for query images. This work provides a robust solution for semantic-based texture retrieval in human–computer interaction scenarios.

Content-based image retrieval (CBIR) is essential for managing and searching massive image repositories across a wide variety of applications. Nevertheless, some traditional CBIR systems exhibit low retrieval accuracy because they use predetermined feature weights, lack semantic gaps, and poorly exploit heterogeneous visual features. To overcome such difficulties, the present study will introduce a multi-feature adaptive CBIR framework that combines deep and handcrafted features using an information entropy-based fusion and a trust-based weighting system. Deep convolutional models, combined with complementary low-level descriptors, are used to extract discriminative features in the proposed approach. A PageRank-based similarity propagation strategy is also used to narrow image ranking by leveraging similarity relationships across the globe. Evaluation is performed using standard retrieval measures, such as Mean Average Precision (mAP), Precision at K, Recall at K, and NDCG. The experimental results show that the proposed approach consistently improves performance across benchmark datasets. The framework boosts mAP by up to 8.6% over traditional fixed-weight fusion methods, while Precision@10 and NDCG@10 increase by 6.2% and 7.4%, respectively. The statistical analysis shows that these improvements are significant at the 95% confidence level, indicating that retrieval behavior is robust and reliable. These findings confirm the efficiency of entropy-driven adaptive fusion and ranking refinement in overcoming the major drawbacks of current CBIR systems, and the suggested framework is appropriate for large-scale image search in practice.

Inderscience is a global company, a dynamic leading independent journal publisher disseminates the latest research across the broad fields of science, engineering and technology; management, public and business administration; environment, ecological economics and sustainable development; computing, ICT and internet/web services, and related areas.

Lifelong Content-based Histopathology Image Retrieval via Bilevel Coreset Selection and Distance Consistency Rehearsal

JOIG leads with cutting-edge research in image, video, and graphics technology - your gateway to innovations enhancing daily life and shaping a smart future. Peer-reviewed articles, published quarterly, illuminate the path to tomorrow's solutions.

To retrieve relevant images from a large database, Content-Based Image Retrieval systems (CBIR) are designed centered on the query image’s visual content. Yet, the existing approaches often struggled with cluttered scenes, complex backgrounds, and the Region of Interest (ROI) in the image. Therefore, to address these challenges, this paper introduces You Only Live Once Version 3 (YOLO V3) and Renormalized Entropy - Gaussian Mixture Model (RE-GMM) approaches. Primarily, the image datasets are collected and pre-processed. The foreground and background objects are separated. Next, by using YOLOV3, the separated objects are detected. Further, by using Gram-Graph Cut (G2C), detected objects are segmented, and saliency mapping is carried out. Afterward, the features are extracted. Then, by using the Gaussian Mutation Tuna Swarm Optimization Algorithm (GM-TSOA), the features are selected from the extracted features. Afterward, by using the AD-ES-CNN algorithm, the object classification is performed. Also, the structural similarity index and semantic feature similarity are carried out by using the SIFT and Jaccard Index, respectively. Lastly, by using the EPL-Fuzzy approach, the object is retrieved and indexed. As per the experimental analysis, the proposed model attained 99.29% accuracy for the Caltech 256 image dataset and 98.5% accuracy for the Corel image dataset.

The subject of the study is the method and algorithms for content-based image retrieval within the Multidimensional Cube (MDC) model. The goal is to develop a search method based on image descriptor vectors and an algorithm that implements this method in both sequential and parallel versions for MDC. The research tasks include: defining requirements for the search method; analyzing the MDC model structure and defining the approach to the search method; developing search methods and algorithms for scenarios where the model is stored in RAM or in a relational database; integrating parallel computing into the algorithm; analyzing alternative models based on multidimensional trees, graphs, hashing, inverted indexing, quantization and inverted multi-index structures; developing evaluation metrics and conducting experiments to compare the efficiency of the MDC-based method with alternative search models. Methodology: analytical and comparative methods for search algorithm evaluation, modeling, and experimental verification were applied. Thread-level parallelism and hardware optimization methods were used, along with comparative analysis of model efficiency (KD-tree, Locality-Sensitive Hashing, Hierarchical Navigable Small World, Inverted File with Flat Compression, Inverted Multi-Index). Statistical methods were employed to assess results using recall, search time, and model construction time metrics. Experiments were conducted with both web-sourced and synthetic image descriptors, as well as load testing to evaluate the model’s throughput. Results: a new search method and the Wave-Search Algorithm were developed. Its parallel version achieves up to a 3x speedup. For top-10 and top-100 queries in a dataset of 1 million descriptors, MDC shows the best overall performance among the compared models based on the metrics and strong stability under load. Conclusions: the proposed search method and its implementation (Wave-Search Algorithm) efficiently utilize the MDC model’s structure for search tasks, outperforms alternative search models in terms of effectiveness, demonstrates robustness under load, and has significant potential for further development, including the use of hardware acceleration.

The rapid growth of digital video data makes efficient Content-Based Video Retrieval ‎‎(CBVR) increasingly important, yet traditional similarity measures often fail to capture high-‎er-order dependencies between video features. This paper introduces a CBVR pipeline that ‎uses a novel non-square determinant kernel as the direct similarity score with a faster Chio-‎like algorithm that reduces matrix order by four in each step. Experiments show an average ‎execution-time decrease of about 25 % compared to the standard Chio-like method and 3.1 % ‎compared to its modified version. Integrating this kernel into the CBVR demonstrates that ‎the Chio-enhanced determinant kernel outperforms similarity measures across benchmark vid-‎eo datasets. By demonstrating superior retrieval efficiency and accuracy, the proposed meth-‎od is well-suited for efficient and accurate similarity evaluation in large-scale or real-time ‎CBVR applications‎.

Finding pictures and movies fast and precisely has become more crucial due to the explosive growth of multimedia content. In order to improve the efficiency and semantic significance of that procedure, this research presents a sophisticated AI-powered retrieval system. The system supports both text-based and image-based searches by combining Facebook AI Similarity Search (FAISS) [10][11]with Contrastive Language–Image Pre-training (CLIP). It produces more accurate results by enabling configurable weighting and removing irrelevant or negatively associated suggestions. During video retrieval, the system extracts individual frames using FFmpeg and indexes them using FAISS for frame-level similarity matching. With Precision@5 of 92.8%, Recall@10 of 89.1%, and an average query time of just 0.45 seconds, the method achieves remarkable performance. Recent developments in multimodal video processing[25],[26] and CLIP optimization[24] are combined to increase efficiency even more. All things considered, our approach offers a scalable and useful foundation for high semantic comprehension in real-time multimedia retrieval.

Content-based mammographic image retrieval requires exact BIRADS categorical matching across five classes, posing far greater complexity than conventional binary classification. Existing studies remain limited by small sample sizes, improper patient-level separation, and inadequate statistical validation, restricting clinical translation. We developed a comprehensive evaluation framework systematically comparing CNN architectures (DenseNet121, ResNet50, VGG16) under advanced training strategies: fine-tuning, metric learning, and super-ensemble optimization. Rigorous patient-stratified splits (1003 patients, two images each), 602 test queries, and bootstrap confidence intervals (1000 resamples) ensured reliable assessment. Advanced fine-tuning and test-time augmentation (TTA) yielded a precision@10 of 34.71% for DenseNet121 _AdvancedFT_TTA, a 25.74% improvement over the baseline ResNet50 (27.6%). Selective super-ensemble and metric learning approaches were further benchmarked under patient-exclusive splits, confirming robust performance across architectures. Statistical analysis (bootstrap CIs, n = 1000; t-tests p < 0.001; Cohen's d > 0.8) validated significant gains and reproducibility. These results establish DenseNet121_AdvancedFT_TTA as the new state-of-the-art for five-class BIRADS retrieval while reducing computational cost.

In the age of digital transformation, effective Image Retrieval (IR) systems are essential for managing the vast amounts of visual data generated daily. Traditional IR methods, primarily reliant to Text Based annotations, face significant challenges, including the labor- intensive process of manual tagging and the inherent subjectivity in human perception. These limitations often lead to inefficiencies and inaccuracies in retrieving relevant images, underscoring the urgent needs for innovating approaches that can enhance retrieval capabilities. This review paper addresses existing gaps in the literature by investigating the role of advanced algorithms in improving image retrieval systems and assessing the impact of external factors on evolving trends. By exploring the integration of multimodal fusion techniques that combine various data sources such as text, images, and audio to enhance the effectiveness of Content Based Image Retrieval (CBIR). Furthermore, identifying ongoing challenges within current methodologies and purpose future research directions that can pave the way for more robust systems. The importance of this review lies in its unique focus on AI driven solutions that leverage Deep Learning (DL) to overcome traditional limitations. By providing clear insights into trends and patterns, this paper aims to highlight the transformative potential of CBIR across various sectors, including e-commerce, digital media, and textiles. Moreover, by emphasizing thee societal relevance of efficient IR such as improving users experience in online shopping demonstrating how advanced CBIR systems can significantly impact everyday life. Ultimately, this paper seeks to contribute to a deeper understanding of how AI can reshape IR methodologies, making that more effective, scalable and user centric in a data driven world.

In response to the demand for quick retrieval from huge picture collections, Content-Based Image Retrieval (CBIR) has grown in prominence as a field of study. A CBIR technique that combines color, texture, and form features is proposed in this paper. By segmenting photos into areas and determining color moments for each, color characteristics may be extracted. Gray-Level Co-occurrence Matrices (GLCMs) are used to assess texture. Five Fourier Descriptors are used to represent shape features. The 1000 photos in 10 categories of the Corel-1k database are used to test the system, which is constructed using MATLAB. Metrics for recall and precision are used to assess performance. The outcomes demonstrate enhanced retrieval precision in comparison to current techniques for all ten image classes. Additional texture and color characteristics may be investigated in future research depending on the application.

Modern technology has made storing, sharing, and organizing huge amounts of data simple through the Internet of Things. Search engines and query-based retrieval databases made access to relevant data easy through ranking and indexing based on stored content. In this paper, a secure CBIR scheme based on watermarking is proposed. Firstly, the image owner embeds the watermark in the image using quantization index modulation (QIM) in the luminance (Y) color space. The watermarked images are then uploaded to the cloud server, which extracts image feature vectors. In this article, features derived from a pre-trained network model from a deep-learning convolutional neural network trained for large image classification have been used for the retrieval of similar images. The image similarity is calculated using Euclidean distance, and the precision (P) is used as the performance measure of the model that achieved nearly 100%. Extensive experiments are carried out, and assessment results reveal the outperforming result of the proposed technique compared to other related schemes. The scheme can be used in many applications that need CBIR, such as digital libraries, historical research, fingerprint identification, and crime prevention.

Leveraging 3DCNN and Weighted Similarity Metrics for Enhanced Content-Based Video Retrieval

The object of the study is the process of organizing a descriptor repository in content-based image retrieval systems. The subject of the study is a method of numerical optimization of descriptor clustering in a multidimensional space. The aim of this work is to develop a clustering optimization method in the Multidimensional Cube model to improve search efficiency. The core idea is to ensure a more uniform distribution of descriptors across clusters by adjusting interval boundaries in each dimension, which reduces imbalance in cluster density and improves retrieval performance. The research methodology combines analytical determination of the number of clusters with numerical optimization of descriptor distribution across intervals. The proposed method, named the Dimension Intervals Numeric Optimization Algorithm, has been implemented in two variants: one for deployment in an external relational database and another for deployment in main memory. Theoretical analysis of computational complexity demonstrated that the proposed approach does not require multiple iterations, unlike the competing methods considered for comparison, namely k-means and the Inverted Multi-Index, and it has lower asymptotic complexity. Experimental evaluation was carried out on a dataset of image descriptors. The results showed that k-means provides the highest clustering quality in terms of cluster density, but requires significantly more time. The proposed method in the main-memory variant demonstrated the best balance between clustering quality and execution time, approaching the quality of the Inverted Multi- Index while outperforming it in runtime. The external-database variant proved slower due to query-processing overheads but is appropriate for scalable systems with centralized data repositories. The conclusion is that applying the developed numerical optimization method enables a more uniform distribution of descriptors across clusters and reduces imbalance in their density.

Leveraging Spatio-temporal Deep Learning and Fuzzy Class Membership for Robust Content-Based Video Retrieval

Abstract The rapid proliferation of video data from various sources underscore the pressing need for effective Content-based Video Retrieval (CBVR) systems. Traditional retrieval methodologies are increasingly inadequate for managing the complexities and scale of video big data, which necessitates the development of advanced distributed computing frameworks. This study identifies and addresses critical challenges in CBVR , specifically the implementation of lambda architecture for the retrieval of both streaming and batch video data, the enhancement of in-memory analytics for video data structures, and the efficient indexing of heterogeneous video features. We propose $$\uplambda \text {-}\hspace{2pt}\mathcal {CLOVR}$$ , a novel scale-out system which integrates state-of-the-art big data technologies with deep learning algorithms. The system architecture is inspired by lambda principles and is designed to facilitate both near real-time and offline video indexing and retrieval. Key contributions of this research include: (1) the formulation of a lambda-style architecture tailored for video big data, (2) the development of an in-memory processing framework that provides a high-level abstraction for video analytics, (3) the introduction of a unified distributed indexer, termed Distributed Encoded Deep Feature Indexer (DEFI), capable of indexing multi-type features from both streaming and batch video datasets, and (4) a comprehensive bottleneck analysis of the proposed system. Performance evaluations utilizing three benchmark datasets demonstrate the system’s effectiveness, revealing insights into performance bottlenecks related to storage, video stream acquisition, processing, and indexing. This research provides a foundational framework for scalable and efficient video analytics, significantly advancing the state-of-the-art in cloud-based CBVR systems.

Introduction: Content-Based Image Retrieval (CBIR) systems aim to retrieve relevant images from large databases based on visual content rather than metadata. Traditional CBIR techniques often struggle with accuracy due to limited feature representation. Recent advances in deep learning offer new opportunities for extracting high-level features. This work explores the integration of deep neural networks into CBIR. Specifically, it utilizes the ResNet50 architecture for enhanced image understanding and retrieval. Objectives: The primary objective is to develop an efficient and accurate CBIR system using deep learning. It aims to leverage the ResNet50 model to extract meaningful image features. The system is designed to improve retrieval accuracy over traditional methods. It seeks to handle diverse image content with flexibility. The goal is to display the top 10 visually similar images for any given query image. Methods: A pre-trained ResNet50 model is used to extract high-level features from a curated dataset of 1,000 images. These features are stored for comparison during retrieval. When a user submits a query image, features are extracted and compared using cosine similarity. This approach combines deep feature extraction with similarity-based ranking for effective image retrieval. Results: The proposed system integrates Deep Learning using CNN and ResNet50 with cosine similarity to perform efficient image retrieval. The model accurately extracts and compares features, ensuring meaningful retrieval results. Figures 3–6 demonstrate sample trials, showcasing the system's ability to retrieve and rank images based on cosine similarity values. Conclusions: This research presents a CBIR system enhanced with deep learning using the ResNet50 model for accurate image retrieval. When a user submits a query image, the system retrieves and ranks the top 10 most similar images using cosine similarity. By leveraging deep learning, the system captures complex visual features, offering improved accuracy over traditional CBIR methods.