一種新的影像驗證方案於電子商務退款應用

科技進步提升影像品質，使得影像編修軟體操作更加簡易，影像完整性成為不可忽視的議題。與此同時，影像品質提升也相對擴增數位媒體的規模，產生傳輸耗時及儲存成本增加的問題。在電商情境中，消費雙方因為退款驗證問題產生糾紛，故若能提供一種有效公正的驗證機制，保護影像不受篡改並保有不錯的品質，可保障消費雙方權益，減少後續糾紛產生。
本研究以解決電子商務退款影像驗證及嵌入影像驗證碼方法的選擇問題為核心，透過文獻回顧，選擇基於AMBTC的影像壓縮技術進行嵌密，設計閾值模型作為兩種嵌密方法的選擇基準，再將求解結果進行比較分析，最終提出本研究之結論與建議。
本研究利用灰階、全彩影像進行模型的可行性測試，研究結果發現除嵌入2個位元及10個位元長度之驗證碼時，閾值有增加的情形，其餘皆與模型相符，產生的影像品質優於過去文獻中僅使用一種嵌密法的結果。此外，我們分別測試嵌入4個位元及8個位元的結果，檢測率為87.85%及94.56%，表示本研究方法可偵測到多數的篡改，保護影像完整性。
綜上所述，本研究的學術貢獻為考慮影像複雜度後提出之轉換閾值模型，可根據嵌入驗證碼的長度提供閾值選擇建議，大幅降低失真並保有良好的可檢測性；實務貢獻則是透過將本研究之技術嵌入攝影應用程式中，可作為電商平台業者之驗證機制，提供店家在短時間內驗證消費者開箱影像之真偽，以保護消費雙方權益減少糾紛。

This study focuses on the selection of refund image verification and embedded image verification code in e-commerce. Through literature review, AMBTC based image compression technique was selected for embedding, and threshold model was designed as the selection benchmark of two embedding methods. Conclusions and suggestions of this study were finally proposed after comparative analysis of the solution results.
In this study, grayscale and full-color images were used to test the feasibility of the model. The results showed that the threshold values of ACs with 2 bits and 10 bits were increased, and other results were consistent with the model, and the image quality was better than that of the previous literature using only one embedding method. In addition, we tested the results of embedding 4 bits and 8 bits respectively, and the detection rates were 87.85% and 94.56%, indicating that this method could detect most tampering and protect image integrity.
To sum up, considering the image complexity, the transformable threshold model proposed in this study can provide a threshold selection basis according to the length of ACs embedded, greatly reducing distortion and maintaining good detectability. By embedding the model into a photography application, the seller can verify the authenticity of the unpackaged image in a short time, so as to protect the rights of consumers and reduce disputes.