| Summary: | 碩士 === 國立臺灣科技大學 === 資訊工程系 === 99 === In recent years, the fields of the image processing applications, such as the digital camera and the image editor, are continually growing and quite important. However, it leads to the problems that digital images may easily be modified and altered. The altered images maybe be infringed upon the property-rights when these images are wrong used. Therefore, we proposed a forensics system to detect the candidate tampered regions which are generated by using the human skin color recognition technique. In this thesis, our proposed method combines two strategies which are the tampered-JPEG-image detection and aliasing edges detection. The first adopted strategy, the tampered-JPEG-image detection, is used to detect the candidate tampered regions according to the double quantization effect. Then, the second strategy, aliasing edges detection, is applied to detect the cropped margins of the candidate tampered regions. Based on the ninety-seven test images in including seventy tampered images and twenty-seven original images, the experimental results demonstrate that the proposed system can accurately detect the tampered images.
|
|---|
