Hybrid CNN–RNN Framework for Intelligent Image Noise Removal and Quality Enhancement
DOI:
https://doi.org/10.65421/jshd.v2i2.157Keywords:
Image enhancement, Image quality assessment, Noise reduction, Convolutional neural networks (CNN), Recurrent neural networks (RNN), Hybrid deep learning model, PSNR, SSIMAbstract
Improving image quality is crucial for accurate analysis and interpretation in applications such as medical imaging, remote sensing, surveillance, and night vision systems. However, images captured in real-world environments often suffer from noise and distortion, reducing their clarity and the accuracy of their analysis. This study proposes a hybrid deep learning model that integrates convolutional neural networks (CNNs) and recurrent neural networks (RNNs) to enhance image quality while minimizing noise and preserving structural details. The proposed model was evaluated using images with varying noise levels and assessed using multiple image quality metrics, including the maximum signal-to-noise ratio (PSNR), structural similarity index (SSIM), mean squared error (MSE), and the learned perceptual image patch similarity index (LPIPS). The experimental results showed significant improvements, with the peak signal-to-noise ratio increasing from 24.85 dB to 33.47 dB and the structural similarity index improving from 0.71 to 0.93, along with substantial reductions in the mean squared error (-73.4%) and the learned perceptual patch similarity index (-66.7%). Statistical analysis confirms the significance of these improvements (p < 0.001). Results showed that the model achieved an accuracy of 95.7%, a recall rate of 94.8%, and an F1 score of 93.2%. Furthermore, significant improvements were observed in image quality metrics, including PSNR and SSIM, with overall improvement levels ranging from 40% to 60% compared to conventional noise reduction techniques. These results demonstrate that the proposed hybrid framework, combining convolutional and recurrent neural networks, offers an effective and robust solution for improving image quality in noisy environments. The results also indicate that the convolutional-recurrent neural network hybrid model provides a powerful and efficient solution for improving image quality in complex real-world environments.
