Handwriting Forgery Detection Using Image Processing

Document forgery detection is a vitally important field because the forensic role is used in many types of crimes because of they are easy to apply and hard to detect.Document examination is an important task to detect the illegal modifications and control the crimes.The most common types of document forgery methods are addition and alteration.Moreover,the forgers use similar inks for modifying the documents,where the human's eye cannot detect these modifications.In this thesis,two methods are proposed to detect the forgery in a text by detecting different ink using image processing instead of conventional methods.The first method is depending on extracting nine features from R,G,and B channels and calculate the distance between these features of all objects of the document.All documents are scanned as an image and segmented into objects.Then nine features are extracted from each object based on R,G,and B channels.Distance measurements between each nearby pairs of feature vectors are computed using root mean square error.Modified Thompson Tau test is applied to extract anomaly points.The tampered points are then obtained exactly from anomaly points.Modified Thompson Tau test has a high detection efficiency and a low omission ratio but its precision is not ideal.Therefore,the second outlier detection has been used to help to make up the difference in precision.The experimental results show that our proposed method can not only detect but also localize tampered objects efficiently.The second method is using Structural Similarity Index of Color Histogram.All documents are scanned as images,then each image is segmented into objects.For each object,color histograms are extracted from R,G,and B channels and feature vector is computed depend on these histograms.Structural similarity index is used to measure the similarity among all objects feature vectors.Finally,the average of similarity values for each object is examined by a threshold to detect the forged objects.The experiment results show that the proposed methods can detect handwriting forgeries and localize the forged objects with high precisions under alteration and addition forgery.