Image Quality Assessment for Map-Merge Quality Evaluation.

In the context of Image Quality Assessment (IQA), our research addresses the challenge of assessing the quality of map-merge applications. Conventionally, human evaluation must be performed to assess the quality, which is time-consuming. Thus, we adopt a strategy that leverages transformer models within the IQA framework to enhance map-merge quality evaluation. We employ a modified transformer model with encoder and decoder networks complemented by a ResNet backbone. The results of our study demonstrate the effectiveness of the proposed IQA model in predicting map-merge images, a prediction that aligns well with human judgment. The PLCC and SRCC metrics of the prediction model are well-correlated with the human opinion score, which lies at 0.89 in the map-merge dataset. Based on the experiment, we successfully established an IQA model tailored for map-merge image applications.