A trust region method for a semismooth reformulation to variational inequality problems