Illuminant-Based Transformed Spaces for Image Forensics

ABSTRACT:

In this paper, we explore transformed spaces, represented by image illuminant maps, to propose a methodology for selecting complementary forms of characterizing visual properties for an effective and automated detection of image forgeries. We combine statistical telltales provided by different image descriptors that explore color, shape, and texture features. We focus on detecting image forgeries containing people and present a method for locating the forgery, specifically, the face of a person in an image. Experiments performed on three different open-access data sets show the potential of the proposed method for pinpointing image forgeries containing people. In the two first data sets (DSO-1 and DSI-1), the proposed method achieved a classification accuracy of 94% and 84%, respectively, a remarkable improvement when compared with the state-of-the-art methods. Finally, when evaluating the third data set comprising questioned images downloaded from the Internet, we also present a detailed analysis of target images.

PROJECT OUTPUT VIDEO:

EXISTING SYSTEM:

• Methods that explore some degree of illumination inconsistencies for detecting image splicing have been the focus of many researchers for over a decade. Basically, they can be divided into two types of approaches: (a) those that look for inconsistencies in light source environment; and (b) the ones that look for inconsistencies in the estimated light source color from the image.

• Wu and Fang have proposed a new way to detect forgeries using illuminant colors. Their method divides a color image into overlapping blocks estimating the illuminant color for each block.

• The authors used the algorithms Gray-World, Gray-Shadow, and Gray-Edge to estimate the illuminant color. In addition, they used a maximum likelihood classifier proposed by Gijsenij and Gevers to select the most appropriate method for representing the illuminant of each block. For forgery detection, the authors choose some blocks as reference and estimate their illuminants.

• Aiming at reducing the user dependency, de Carvalho et al. proposed a different way to use illuminants for detecting forgeries. In a custom-tailored method for detecting image compositions involving people, the authors estimate illuminant maps for the image using statistics-based and physics based approaches.

DISADVANTAGES OF EXISTING SYSTEM:

• The problem with this method is the need of well-defined specular highlight regions for estimating the illuminants.

• Unfortunately, the high degree of user dependence for the selection of superpixels and for the distance map analysis makes the method strongly susceptible to human errors.

PROPOSED SYSTEM:

• In our approach, inconsistencies of color, texture and shape present in a fake image become more pronounced in the transformed image, which is obtained converting an input image to different illuminant maps.

• More specifically, this work extends upon the method recently proposed by de Carvalho et al. in which the authors use texture and edge descriptors to characterize IMs and detect inconsistencies in images pointing out possible tampering operations.

• In this work, we extensively study different ways to use combinations of different IMs for different color spaces and examine the most appropriate image descriptors and classifiers to better capture visual properties that might lead to forgery detection. We strive for exploring complementary features in order to achieve a very effective classification rate in different setups.

• Our main contributions herein are:

• The use of color descriptors computed upon transformed image spaces (illuminant maps, IMs) and a full study of the effectiveness and complementarity of these image descriptors computed on such transformed spaces;

• The adoption of a machine learning framework in the proposed approach, for automatically selecting the best combination of all the factors of interest (e.g.,transformation spaces (IMs), color-space representations, descriptors, and classifiers;

• A quantitative evaluation of the differences among pristine and fake images when represented in different IM spaces; and

• The introduction of an approach to detecting the most likely doctored part in fake images.

ADVANTAGES OF PROPOSED SYSTEM:

• To decrease user interaction and increase the classification accuracy on image splicing detection.

• The proposed method yields an improvement of 15 percentage points in the classification accuracy and the possibility of providing a confidence degree associated with the classified image when compared to the state-of-the-art.

• The proposed system presented an improved approach to detecting composites of people that explore complementary information for characterizing images.

MODULES:

1. Image Acquisition
2. Preprocessing
3. Feature Extraction
4. Forgery Classification
5. Forgery Detection

MODULE DESCRIPTION:

The splicing detection process commonly relies on the expert’s experience and background knowledge. This process usually is time consuming and error prone as image splicing are ever more sophisticated, and an aural (e.g., visual) analysis may not be enough to detect forgeries.

Our approach to detecting image splicing, which is specific for pinpointing composites of people, is developed aiming at minimizing the user interaction. The splicing detection task consists in labelling a new image among two pre-defined classes (real and fake) and later pointing out the face with higher probability to be the fake face.

Image Acquisition:

• The first stage of any vision system is the image acquisition stage.

• The image has been obtained by Gallery.

Preprocessing:

Preprocessing having two steps,

1. Illuminant Map Extraction

2. Face Detection

i. Illuminant Map Extraction:

We proposed a new way to detect forgeries using illuminant colors. Their method divides a color image into overlapping blocks estimating the illuminant color for each block. Many algorithms such as Gray-World, Gray-Shadow, and Gray-Edge are used to estimate the illuminant color. In our system estimate illuminant color using Gray-World [statistical-based information extracted].

ii. Face Detection:

After estimate the illuminant color in image, detect face one by one using face detection tool. Then detected each face images cropped for further process.

Feature Extraction:

In feature extraction stage, cropped illuminant map face images are processed. In this stage image descriptor such as Content–based image retrieval method is used for feature extraction. Color, Texture and shape features are extracted in cropped illuminant map face images. Finally encoding the extracted information into feature vectors for further process.

• In color based, color HSV histogram, autocorrelation and color moments features are extracted.

• In texture based, energy, contrast, correlation and homogeneity features are extracted using GLCM method.

• In shape based, edge feature are extracted.

Forgery Classification:

In this stage, faces are classified in the form of real or fake based on color, texture and shape features.

Forgery Detection:

In this stage, once knowing that an image is fake, this stage aims at identifying which face is more likely to be fake in the image using surf feature detection.

Speeded Up Robust Features (SURF) is a local feature detector and descriptor. It that can be used for tasks such as object recognition, image registration, classification or 3D reconstruction.

SYSTEM REQUIREMENTS:

HARDWARE REQUIREMENTS:

• System            : Pentium Dual Core.

• Hard Disk       : 120 GB.

• Monitor          : 15’’ LED

• Input Devices : Keyboard, Mouse

• Ram               : 1GB.

SOFTWARE REQUIREMENTS:

• Operating system : Windows 7.

• Coding Language : MATLAB

• Tool                      : MATLAB R2013A

REFERENCE:

Tiago Carvalho, Fábio A. Faria, Hélio Pedrini, Senior Member, IEEE, Ricardo da S. Torres, Member, IEEE, and Anderson Rocha, Senior Member, IEEE, “Illuminant-Based Transformed Spaces for Image Forensics”, IEEE TRANSACTIONS ON INFORMATION FORENSICS AND SECURITY, VOL. 11, NO. 4, APRIL 2016.