この資料をSMS送信: Mitigating Object Hallucinations in Large Vision-Language Models through Visual Contrastive Decoding