Application of local invariant symmetry features to detect and localize image copy move forgeries

Abstract

Today whatever is seen in digital images could be unrealistic due to the advent of sophisticated systems and image editing software's. It is easy to edit digital images without leaving the traces of manipulation, therefore tampering is hard to discern visually. Among the types of digital image forgeries, realizing the copy move forgery is very challenging. Hence, this paper proposes a novel scheme to detect copy-move forgery by means of symmetry based local features. The proposed scheme can also detect the multiple copy move forgeries and localize the detected regions. The experiments are carried out by using the various datasets, and comparative study also made with the existing methods.

Introduction

In today's world, images are stored in a digital form which carries vital information and hence the integrity of digital images is very essential. For example, digital images can be used in the day to day life in medical diagnosis, television news, magazines, newspapers, and as legal evidence, etc… These images can be easily accessed and their content can also be tampered with using advanced image editing software. In the past decades, the forgery has been detected using the techniques of image watermarking [1], [2], [3], [4], [5] and digital signatures [6], [7], [8], [9] and these techniques require some external information such as watermark and hash value. On the other hand, the latest research studies focus on detecting the image forgeries without external information [10], [11].

There are many kinds of image forgeries and among them, the “copy move/copy paste/cloning forgery” is a challenging one to detect because this forgery is done by copying a portion of the image and pasting it in another portion of the same image. Moreover, the pasted portion also has similar patterns of the copied portion [12]. This kind of forgery can be done either for the purpose of concealing some information or to raise the number of objects. The same has been pictorially presented in Fig. 1, where, a truck in the first row of column (a) has been concealed by copy pasting the region contained the leaves and is shown in the first row of column (b). Similarly, the second row of column (a) contains only one ball, which has been copied and pasted in the same image in another portion to increase the number of objects and it is shown in the second row of (b). Such kinds of tampering are identified in this work by finding the matches among the extracted novel local symmetry features. As the symmetry features are potentially stable and robust when it is explored at multiple scales and multiple locations, local symmetries also can be relatively descriptive. The numerous work has been previously implemented are based on the SIFT (Scale Invariant Feature Transform) and SURF (Speeded Up Robust Features) features. The main disadvantage of SIFT features is a high computational cost, especially in the process of feature extraction and matching. Even though the SURF features are an efficient one, but it lacks in finding the symmetrical pairs of feature points. So, the proposed scheme proposes a novel scheme to extract such features by computing the local symmetry value; these features are computed by analyzing the differences over the symmetry axes. The proposed scheme is designed an algorithm to detect one than one copy-move forgeries and multiple copies of forgeries in a single attempt.

The major objectives of this research study are, (i) to propose a new local invariant feature to detect the forgery, (ii) to design an algorithm to detect single and multi copy move forgeries in a single attempt and (iii) to evaluate the performance of the proposed scheme. The remaining of the paper is structured as follows: Section 2 states about the previous methods, Section 3 furnishes the proposed scheme, Section 4 discusses the experimental results and finally, Section 5 gives the conclusion of the paper.