To satisfy the testing requirements for interactions between autonomous vehicles (AVs) and pedestrians, this paper presents a test scenario library generation framework adaptable to different operational design domains (ODDs), AV models, and performance metrics. The framework constructs virtual test scenarios focusing on vehicle-pedestrian interactions, increasing the number of critical test scenarios. In the scenario generation process, we first extract raw scenario data from naturalistic driving datasets involving interactions between vehicles either proceeding straight or turning at signalized intersections and pedestrians crossing the road. Based on the key characteristics of these vehicle-pedestrian interactions and considering the occurrence probabilities of the scenarios, we establish a driving safety field model to characterize the interaction risks between vehicles and pedestrians. Finally, we propose a criticality function that combines exposure frequency and danger, and design an auxiliary objective function. By using an optimization algorithm to search for critical scenarios, we enhance testing efficiency and achieve accelerated testing.