Perceptual Quality Restoration of Laparoscopic Videos via GAN-Based Smoke Elimination
Akib Jayed Islam, Sultanus Salehin, Sayem Ul Alam, Shreya Paul · 2024
A supervised image-to-image translation method is developed to remove smoke from laparoscopic video recordings, which can be significantly damaged due to laparoscopic surgery. Tissue dismemberment tool smoke can obscure a surgeon’s view and cause mistakes in computer vision-based systems that are utilized for clinical navigation operations. Redesigned Generative Adversarial Network (GAN) is introduced to eliminate the noise while maintaining an image quality that is perceptually adequate. The proposed approach is trained and tested using the Laparoscopic Video Quality (LVQ) assessment dataset. The experimental result segment provides examples of the qualitative results obtained using the proposed approach. More refinement of the results is anticipated if a bigger and more comprehensive database is used for testing and training. The recommended method outperformed existing methods and generated smoke-free photos or videos.