Online hashing has attracted much research attention for large-scale image retrieval in a streaming way. The main challenge lies in keeping balance between high retrieval accuracy and low training time. Existing online hashing methods almost rely on shallow models rather than deep networks due to high training costs, because it is unacceptable to update hash functions on an order of hours. In addition, the multi-label supervision information is not fully utilized to guide the hash learning process and the affinity matrix is always fixed once constructed. In this paper, we propose a novel Deep Graph Online Hashing (DGOH) method, which for the first time introduces inductive graph neural networks (GNNs) to realize deep online hashing with acceptable training costs on an order of seconds. Furthermore, we mine the multi-label information of the images by constructing a label network and learn label-wise weights dynamically to help to update the affinity matrix. In addition, we provide a strategy to obtain examples from the old data to solve the catastrophic forgetting problem. An integrated objective function is designed to train the entire architecture. Extensive experiments on two common benchmarks demonstrate that the proposed method achieves up to 13.3% accuracy gains over state-of-the-art baselines and shows competitive performance on training time.

With the growing popularity of cloud services, companies and individuals outsource images to cloud servers to reduce storage and computing burdens. The images are encrypted before outsourcing for privacy protection. It has become urgent to solve the privacy-preserving image retrieval problem on the cloud. There are three main challenges in this area. First, how can we achieve high retrieval accuracy on the encryption domain? Second, how can we improve efficiency in large-scale encrypted image retrieval? Third, how can we ensure the reliability of the retrieval results? The existing schemes only consider some of these characteristics and the retrieval accuracy is insufficient. In this paper, we propose a privacy-preserving large-scale image retrieval framework with vision graph convolutional neural network hashing (ViGH). To the best of our knowledge, this is the first framework that is able to address all the above challenges with more advanced accuracy performance. To be specific, cycle-consistent adversarial networks and vision graph convolutional networks (ViG) are utilized to increase retrieval accuracy. By embedding encrypted images into hash codes, we can obtain high retrieval efficiency by Hamming distances. Cloud servers store the hash codes on the blockchain (Ethereum). The retrieval algorithm on the smart contracts and the consensus mechanism of blockchain ensure reliability of the retrieval results. The experimental results on three common datasets verify the effectiveness and efficiency of the proposed privacy-preserving image retrieval framework. The reliability of the retrieval results is ensured by the consensus mechanism of blockchain with no need for verification.

Deep hash networks have garnered significant attention due to their efficiency and ability to learn discriminative embeddings for approximate nearest neighbor search. However, it is observed that deep hash networks are vulnerable to adversarial interference, which is an important security problem. Despite the growing interest in targeted attack on deep hash networks, it suffers from a scarcity of research on generating universal adversarial perturbations which are unrelated to the specific images. In this paper, we introduce a novel Targeted Universal adversarial Attack (TUA) on deep hash networks. Our framework consists of two key components: a ReferenceNet and a universal generative adversarial network. Specifically, ReferenceNet is designed to generate category-level representative reference codes for the target labels by introducing a cosine similarity based reference loss. Additionally, we feed the fixed random noise and target labels into the generator to learn universal adversarial perturbations. Particularly, the reference codes are used to optimize the generator by minimizing the Hamming distances between the hash codes of the adversarial examples and the reference codes. Extensive experiments on three common datasets validate the superior targeted attack performance, transferability, and universality of our method compared with state-of-the-art targeted attack methods on deep hash networks.

Hashing is frequently used in approximate nearest neighbor search due to its storage and search efficiency. On account of the bottleneck of traditional learning to hash methods, deep-based learning to hash has gained quite a popularity among researchers recently. While such methods show a promising performance gain by utilizing deep neural networks into its end-to-end training process to generate compact binary codes, the intrinsic connections between components make it unfeasible to optimize the architecture significantly. Subject to noise interference and absence of similarity labels of training data, normal unsupervised deep models even carry a noticeable deviation at representation learning stage. By integrating Generative Adversarial Network, this paper presents a novel architecture for generating compact hash codes from an extended set derived from original images while using the benefit of similarity prediction. Extensive experiments conducted on three benchmarks, NUS-WIDE, CIFAR-10, and MS-COCO illustrate that our method outperforms state-of-the-art image retrieval models and generates high-quality binary hash codes seamlessly.

With their powerful feature extraction ability, convolutional neural network (CNN) models achieve excellent signal reconstruction and recovery performances compared with those of traditional methods. The CNN-based approaches mainly use supervised learning approaches; thus, they require large numbers of ground-truth labeled samples. However, in the seismic denoising field, collecting large numbers of labeled samples is impossible; thus, the main challenge to using deep learning methods is a lack of labeled data. Moreover, the data that are available contain noise. To resolve these shortcomings, this letter proposes a novel self-supervised learning framework to reconstruct and perform blind denoising of seismic data images; this approach requires no labeled training data. We utilize a masking procedure to modify an observation input to a CNN to create a$\mathcal {J}$-invariant function and incorporate a specific CNN architecture known as U2Net, which implements a two-level nested autoencoder that extracts complex feature information from different scales. We modify the network to make it more suitable for seismic signal reconstruction. Finally, we use the self-supervised loss between the original observation and the net output to update the weights of U2Net through backpropagation. Tests on both synthetic and field data demonstrate the superior performance of our algorithm on low signal-to-noise ratio data.

The fractal conservation law (FCL) is a partial-differential-equation-based filtering approach. Analysis of the frequency response of the FCL indicates that it can eliminate high frequencies and preserve or amplify low/medium frequencies. Generally, the shape of the frequency response is fixed. Thus, the FCL cannot track the signal beyond the threshold, which corresponds to the cutoff frequency. Besides, the FCL recovers seismic events only along the time direction, thereby ignoring the coherence between neighboring traces. To resolve these shortcomings, this letter presents a novel spatiotemporal adaptive FCL method. We use a group of frequency responses of the FCL determined by different parameters to construct a convex hull of the filtering results. Then, we introduce an objective function based on the penalized least squares criterion with respect to FCL estimation on this convex hull and take the directional derivative as the penalty term. Thus, the correlations between the adjacent channels are taken into account in the algorithm. Therefore, the 2-D adaptive FCL is equivalent to a convex optimization problem with box constraints, which can be solved using the projected gradient descent algorithm. The application of gradient descent consists of taking the derivative of an objective function, which can be implemented quickly by means of the discrete cosine transform (DCT). The experimental results illustrate that our proposed algorithm has a higher output signal-to-noise ratio (SNR) than the 1-D adaptive FCL on some synthetic records and field seismic data.

In this letter, we have applied a new filtering method for the simultaneous noise reduction and enhancement of seismic signals using a fractal conservation law, which is simply a partial differential equation (PDE) modified by a nonlocal fractional antidiffusive term of lower order. By using an integral formula, which is based on Taylor-Poisson's formula and Fubini's theorem for an antidiffusive term, and then taking the fast Fourier transform of the PDE, the filter in the frequency domain can be derived. The study showed that this filter eliminates the high frequencies, amplifies the medium frequencies, and preserves the low frequencies. Thus, some features of the signal, such as relative maxima or minima, can be preserved or even amplified. Usually, these features are flattened by other denoising methods. The properties of this novel method are tested on both synthetic seismic data and real common shot point seismic records. Additionally, we compared this method with time-frequency peak filtering (TFPF), which has a good performance under low signal-to-noise ratio. The experimental results illustrate the superior performance of our method versus TFPF in the recovery of seismic events by removing random noise and enhancing valid signal.