Additive deep feature optimization for semantic image retrieval

An image retrieval system is required to provide high accuracy in a short time. Combining various features will usually increase accuracy but also increase retrieval time. This study developed a CBIR (Content-Based Image Retrieval) method based on hierarchical clustering on low-level features. Low-level features consisting of color, texture, and shape are extracted and then clustered hierarchically. The resulting clusters are then validated to obtain their optimal number. In the retrieval process, the query image features are extracted and compared with the cluster centroid on each feature. The scores of query results on each feature are normalized, and then the normalized scores are weighted to get the total score. The experiment was carried out using three datasets, namely DIKE20, Corel-1k, and Corel-10k. Based on the experimental result, the proposed method shows better performance compared to the existing state-of-the-art method. On the Corel-1k and Corel-10k datasets, the proposed method obtained precision scores of 0.81 and 0.62, respectively.

Section: Benchmarkingmentioning

confidence: 84%

“…The study by [17] presents a deep convolutional neural network-based model called MaxNet for content-based image retrieval. The proposed MaxNet model contains a total of twenty-one convolution layers that are iterated in a structured way to extract maximum information from the images.…”

Section: Literature Reviewmentioning

confidence: 99%

A Robust Image Retrieval Method Using Multi-Hierarchical Agglomerative Clustering and Davis-Bouldin Index

2022

“…Authors in [18] present a deep convolutional neural network-based model called MaxNet for content-based image retrieval. The proposed MaxNet model consists of twenty-one convolution layers that are iterated in a structured manner to extract the most information from the images.…”

Section: Related Workmentioning

confidence: 99%

“…On the experiment on the Corel-10k dataset, the mean average precision and recall percentage scores were also evaluated on the number of retrieval L images of 10, 20, 30, 40, 50, and 100. Table 6 shows the comparison between the proposed CCBIR method and other methods [17][18][19] on the Corel-10k dataset in terms of precision and recall scores at the top 20 retrieval levels. Authors in [19] 46.2 9…”

Section: The Experiments On the Corel-10k Datasetmentioning

confidence: 99%

Unsupervised Content Based Image Retrieval Using Pre-Trained CNN and PCNN Features Extractors

2023

Content-based image retrieval (CBIR) is a popular approach for searching and retrieving digital images from labeled or unlabeled image collections using image content features. The most important parts of the CBIR system are the computational complexity and the retrieval accuracy. Many studies have been conducted to increase the accuracy of image retrieval systems. In this paper, we propose a new fully unsupervised content-based image retrieval (CBIR) technique to increase the efficacy of image retrieval using a clustering approach on unlabeled image dataset. The proposed method combines features extracted from both pre-trained AlexNet model and pulse-coupled neural networks (PCNN) to extract high and low level features. Then principal component analysis (PCA) is performed on AlexNet's features and these combinations are fed to the K-means algorithm after normalization process. Then Euclidean distance is used to measure the similarity between query and stored images within the same cluster. Finally top similar images are ranked and retrieved. Experimental results on the benchmark Corel-1k and Corel-10k datasets show that the proposed method achieves high precision values of 92.8% and 64.8% respectively, on the top 20 retrieval levels compared to other methods.

“…It is crucial to emphasize that the majority of lower-level features only extract local data and are unable to capture the semantic correlations required for secure CBIR methods. High-level semantic features and fused features have gradually replaced CBIR based on low-level features [13][14][15][16]. Since the introduction of deep convolutional neural networks (CNN).…”

Section: Introductionmentioning

confidence: 99%

Securing Privacy: Encrypted Image Retrieval with CNNs and Chaos-Based Visual Cryptography on Cloud Computing

2023

Data security and accessibility in a cloud computing environment are successfully guaranteed by the safe privacy content-based image retrieval (CBIR) approach. However, ineffective cipher text methods and featureextraction techniques might reduce performance. Therefore, in order to create secure privacy CBIR, it is still necessary to address the need for high-security encryption techniques and the capacity to extract characteristics from cipher text images. To achieve this, we provide a secure privacy image retrieval method based on convolutional neural network features. Initially, a 3D Lotka_Volterra chaotic systems encryption method based on visual cryptography encodes images, and the updated DenseNet-121 model is fine-tuned by employing the encrypted images to construct a feature extractor. There are multi-authorized users set up for the fine-tuning feature extractor. The encrypted features and images are uploaded to a cloud server. Experiments on the Corel10k dataset showed the effectiveness of the proposed model in terms of security and precision compared to previous works, achieving an average search precision of 68.94% and an MSE =12764.27263, PSNR = 7.300480125 dB between the original image and encrypted, and the decrypted image is entirely identical to the original image.