With the extensive use of editing software,digital image tampering be comes very easy,where the image splicing is the most common.Meanwhile,some images such as evidence in co urt are very important.It’s crucial to provide some reliable methods to identify whether an image has been f orged.An image splicing detection method combining run-length with steganography analysis feat ure is proposed.First,the steganalysis feature is extracted by applying a submodel named “s2_spam12hv” ( a part of the rich model proposed by Fridrich et al) into the coefficients matrix generated by block discrete cosi ne transform (BDCT) of an image.Then, this feature is combined with the run-length feature.The run-length feature c onsists of four gray level run-length run-number vectors extracted in four different directions from a de-correlated image.The feature extractions of the two parts are both carried out in chroma space which consists of cb and cr chann els.Support vector machine is chosen as the classifier.Experimental results show that the merged feature can achieve accuracies of 98.57% and 97.27% in datasets CASIA v1.0and CASIA v2.0,respectively.The recognition rate of the feature without merging is greatly improved,and the proposed feature fusion algorithm also shows good recog nition performance compared with some existing algorithms.