Utilizing Vision Large Language Models for Automatic Image Annotations: A Comparative Study