| | 講 題 | A Case Study for Improving Median Edge Detection Predictor on Reversible Data Hiding | | | 講 者 | Dept. of Mathematics, National Changhua University of Education-Hui-Shih Leng | | | 日 期 | 2014/10/24 | 長 度 | 00:13:37 | 人 氣 | 392 次 | | | Recently, many researchers focus on prediction-based reversible data hiding. An important issue is to choose a suitable predictor in order to increase payload and reduce distortion after embedding phase. The median edge detection predictor is a well-known predictor currently used as a JPEG-LS standard. It uses three-sided information and raster-scan order to estimate the target pixel's prediction value and divides the prediction types into two cases: “edge” and “non-edge”. According to the properties of a natural image, in which most pixels are similar to their surrounding pixels in the smooth area. In this study, we focus on “non-edge” case and claim that if the difference value between the horizontal neighboring pixel and the vertical neighboring pixel is less than a threshold, we denote the predicted value of the target pixel as the average value of sum of the three-sided pixel's value. The threshold can be found by employing an exhaustive method from 0 to 255, and satisfy two conditions: smaller mean squared error and has maximum peak value when prediction error is zero. The experimental results show that the proposed method has improved the prediction error of the target pixels to increase payload and reduce distortion of the stego image. | | | 提 供 | TANET台灣網際網路研討會-TANET2014 | | |
|
