摘要
                
                    针对目前图像复制粘贴检测算法有时间复杂度高、当图像中有光滑区域时错检率大的缺点,将图像角点区域及其八方向邻接圆形区域作为伪造检测区域,提取每个检测区域的矩特征后用k-d树进行排序,并通过比较每个区域矩特征的相似性来定位伪造复制区域。该算法效率高,且对经过旋转、添加噪声、模糊、JPEG压缩等后处理的复制粘贴伪造图像检测效果很好。
                
                Many proposed copy-move forgery detection methods are time consuming, and most of them will be ineffective when there are many smooth regions in a forgery image. This paper used the circular regions around image corners as detection regions, and sorted moments of each detected region by using of k-d tree. It could locate the duplicate regions after comparing the similarities of all detection regions' moments. Several experiments demonstrate the efficiency of this method, and also evaluate the copy-move forgery image detecting method' s robustness and sensitivity when a forgery image undergoes a serial of image processing such as rotating, adding noise, blurring, compressing etc.
    
    
    
    
                出处
                
                    《计算机应用研究》
                        
                                CSCD
                                北大核心
                        
                    
                        2011年第12期4782-4785,共4页
                    
                
                    Application Research of Computers
     
            
                基金
                    西北工业大学基础研究基金资助项目(2009-2011)
            
    
    
    
                作者简介
高芳(1987-),女,山西太原人,硕士研究生,主要研究方向为语音图像处理与媒体安全(floraaaaat@126.com);
郑江滨(1971-),男。陕西西安人,教授,主要研究方向为语音图像处理与计算机视觉;
蔡里宁,男,主要研究方向为语音图像处理与计算机视觉.