Comparative Analysis of Subjective Evaluations for Traditional and Neural-Based Video Enhancement Techniques

This work evaluates the effectiveness of modernvideo restoration methods, contrasting neural network-basedtechniques with traditional statistical algorithms to improveperceived video quality. Our analysis focused on three distinctmethods: VBM4D, CVEGAN, and Ramsook, assessing theirperformance using pairwise subjective assessments with a compressed baseline. Results indicate a significant disparity betweenobjective and subjective evaluations, with traditional methods likeVBM4D showing limited improvements in perceptual quality, asdemonstrated by a statistically non-significant increase in MeanOpinion-Score (MOS). In contrast, the neural-based methods,CVEGAN and Ramsook, showed statistically significant improvements in subjective video quality. The findings highlight thesuperior capability of neural approaches to enhance perceptualquality, suggesting that current objective metrics may not fullycapture quality as perceived by human observers. This studyalso contributes the results of the comparative analysis and thedataset to the research community.