Detecting moiré patterns in digital images is essential as it offers insights for assessing image integrity and undertaking demoiréing processes. MoireDet is a simple and efficient moiré pattern detection neural network designed explicitly for moiré edge map estimation [23]. However, the random feature mapping of the Performer significantly affects the prediction for continuous video frames. This paper introduces MoireDet+ based on existing work, which introduces Vision Transformer into moiré-related tasks. MoireDet+ utilizes a mixed-encoder as a backbone, integrating both high- and low-level vision encoders in an FPN-like method, along with a spatial encoder to extract the complex spatial features of moiré patterns. Furthermore, we produce a rapid approximation-based evaluation method to aid loss function design in image restoration and similar tasks. MoireDet+ reaches a state-of-the-art level on mainstream datasets, reducing time costs by 18% compared to MoireDet and other baselines with comparable performance levels.