中圖分類號：TP751.1
文獻標識碼：A
DOI：10.19358/j.issn.2096-5133.2020.02.003
引用格式：劉進林，李欣竹，蔣晨琛.基于改進馬爾科夫特征的圖像拼接檢測研究［J］.信息技術與網絡安全，2020,39（2）：13-18.

Research on image splicing detection based on improved Markov

Abstract： Aiming at the problem that the splicing detection accuracy of traditional Markov feature is not high,this paper proposes an efficient Markov feature extraction method for image splicing detection.Different from traditional Markov feature,this paper only calculates the transition probability matrix in both the horizontal and vertical directions,and chooses the sum of four Markov transition probability values at the corresponding position in the probability matrix as a feature vector.The summation operation not only reduces the number of features,but also enlarges the discrimination of the probability distributions between the authentic and the spliced images.A fixed number of features,regardless of the test datasets,are used in the proposed algorithm. This method achieves the accuracy of 94.38%,99.19% and 96.02% at Columbia image splicing detection evaluation dataset，CASIA V1.0 and CASIA V2.0 detection evaluation dataset respectively.

Key words : image forensics;splicing detection;Markov transition probability;Markov feature selection