Enhancing No Reference Laparoscopic Video Quality Assessment with Evolutionary ANFIS

Abstract Distortions in laparoscopic videos affect surgeon visibility and surgical precision, underscoring the need for sustained high video quality. This study presents a real-time laparoscopic video quality assessment algorithm independent of reference content availability. Statistical parameters derived from luminance, local binary pattern and motion-vector maps of video frames are observed to effectively discern distortion types and severities. These parameters are used to train an evolutionary adaptive neuro-fuzzy inference system (ANFIS) end-to-end with subjective score labels. Training and validation loss curves saturate at the 85 th epoch, demonstrating the model’s efficient data fitting capability. Performance comparison with other state-of-the-art methods reveals superior results, with high correlation scores of 0.9989 and 0.9446 for experts and 0.9956 and 0.9847 for non-experts, alongside low root mean square errors of 0.0828 and 0.1685 for expert and non-experts, respectively. The model accurately replicates the expert and non-expert perceptual opinions, encouraging future research in stereoscopic, augmented, and virtual reality data.