Novel post-photographic technique based on deep convolutional neural network and blockchain technology

【Author】 Geng, Hongjie; Zhou, Mingming

【Source】JOURNAL OF SUPERCOMPUTING

【影响因子】2.557

【Abstract】This work aims to explore the fusion of deep learning and blockchain technology for research applications in photography and art studies. A digital image watermarking model is established, followed by the development of a versatile watermarking algorithm model using singular value decomposition (SVD) and deep learning techniques, thereby extending the applicability of watermarking technology. Furthermore, the performance parameters of the blockchain model are designed using the practical byzantine fault tolerance consensus mechanism to ensure transaction consistency and reliability. Finally, the algorithm's normalized cross-correlation is analyzed and attack experiments are conducted on the original images. Adjustments are made to the model to simulate the scalability of sharded blockchains, and scalability of cumulative revenue is evaluated through simulation experiments. Through simulation experiments, it is demonstrated that this model can capture value functions in a sharded blockchain environment, accelerating convergence speed and achieving scalability. After the optimal value function is obtained, cumulative revenue remains consistently stable, indicating the model's robustness and performance. Deep learning networks are currently successfully extracting key features from photographic works, thereby demonstrating the model's effectiveness in image processing. Attack experiments currently show that 3.901% of the detected images contain collagen. The degree of decentralization has significantly increased after optimization, with a value of 0.896 before optimization and 0.216 after optimization. Post-optimization, no single entity currently dominates the system. The current decentralization degree significantly exceeds the pre-optimization state (P < 0.05), suggesting the model's robustness against anomalous or malicious inputs. The combination of deep convolutional neural networks with SVD and blockchain technology is currently exhibiting strong convergence performance and the ability to extract critical image features during photo post-processing. It currently demonstrates excellent convergence and holds significant potential for applications such as image copyright protection and enhancing image processing in various domains.

【Keywords】Deep learning; Blockchain technology; Post-photographic technology; Neural network; Art research

【发表时间】2023 2023 OCT 17

【收录时间】2023-10-31

【文献类型】实验仿真

【主题类别】

区块链技术-协同技术-深度学习