Correcting the Lens and Purifying the Image: A Deep Learning Approach to Optical Distortion Correction and Image Denoising in Oil and Gas Computer Vision Applications

Abstract

Computer vision technologies are increasingly used across Oil and Gas operations to support inspection, monitoring, and surveillance tasks. In industrial environments, however, image quality is often degraded by lens-induced optical distortions and environmental noise arising from low illumination, dust, vibration, and sensor limitations. These degradations reduce the reliability of visual data and adversely affect the performance of downstream computer vision algorithms. This paper aims to develop and evaluate a unified deep learning-based framework that jointly addresses optical distortion correction and image denoising for Oil and Gas computer vision applications. A convolutional neural network (CNN) is trained in an end-to-end manner to learn mappings from distorted and noisy images to high-quality restored outputs, using datasets collected from representative industrial scenarios. The network incorporates skip connections to facilitate residual learning and is optimized using a hybrid loss function combining mean squared error (MSE) and structural similarity index (SSIM) to preserve both pixel-level accuracy and structural consistency. Experimental results demonstrate that the proposed approach consistently outperforms traditional correction and denoising methods, achieving an average PSNR of 23.46 dB and SSIM of 0.876 on 804 degraded-clean image pairs. The framework further demonstrates strong robustness and generalization across diverse operating conditions, surpassing conventional filtering and geometric correction techniques. These improvements enhance the reliability of vision-based inspection systems and support more robust automated monitoring in harsh industrial environments.