부산대학교 도서관

Digital images are used everywhere in our daily lives. However, because digital images can be edited easily by general users, they need authentication. A method to obtain a morphological pattern spectrum-based total pixel value for objects equal to the structuring element size is proposed to detect image manipulation. The proposed method extracts objects equal to the structuring element size from an image, and the pixel values for the extracted objects are summed. Therefore, the proposed method can detect multiple manipulated areas in an image. This paper shows the results of original and manipulated grayscale images that have large and small manipulated areas. The method proposed in this paper processes 256 × 256 pixel grayscale images with an initial square structuring element with a size of a square measuring 3 × 3 pixels. From these pattern spectrum results, manipulated images are identified. Also, even if the image has a small manipulated area, the total pixel value changes more significantly in each structuring element. Therefore, the morphological pattern spectrum-based total pixel values for objects equal to the structuring element size can detect manipulation because the pixel values for the manipulated image change. Future work is required to confirm the proposed method on manipulated and compressed images to extract objects equal to the structuring element size.