Overview of functionality

Functionality of the application is divided into three modules. This division is highlighted in color in the main menu.

Schéma aplikace Pizzaro

Image source determination

This module provides functions for detecting which camera or video camera was used to take the digital photo or video in question.

Assignment to the source device identifies a particular device (camera or video camera),  which was used to record the data. The functionality includes estimation of noise characteristics of the device (with the possibility of storing in a database), and comparison of the characteristics with the input image.

The process of identifying a particular camera


LCD image re-capture function identifies whether the digital photo comes from a real scene or was captured from a LCD monitor or was scanned with a scanner.

Quantization tables help to find a camera model (or manufacturer) of a device,  which was used to take the digital photo or video. It uses EXIF information and extracts quantization tables used for compression of photos, and searches in a large database of quantization tables and camera models.

Image content verification

This module provides functionality for detecting authenticity of photos and marking any post-processing on a computer after shooting.

Detecting double compression shows repeated use of JPEG compression to the same image. Positive finding indicates that the photo was (after the first compressing and storing in the camera) re-opened and saved by another program. This indicates a high probability of image manipulation.


input data

DCT [0,0]: (left) original histogram, (middle) 2x JPG 85-75, (right) 2x JPG 85-95 

DCT [1,1]: (left) original histogram, (middle) 2x JPG 85-75, (right) 2x JPG 85-95

Detection of interpolation finds traces of interpolation in areas where the image size has been increased or decreased(zoom); It can be applied also to the entire image.

Vstupní obrázek, levý žralok je zvětšený Výstup pro nemodifikovaný objekt Výstup pro zvětšený objekt
input image, left shark is magnified result in the unaltered area (red rectangle)  result in the altered area (note distinctive peaks)

Detection of copy-move function identifies duplicated areas in images. The duplicated areas may result from the use of "clone stamp" tool provided by many graphical editors (repainting of an object, or removing an object).

Původní autentický obrázek Obrázek s vymazaným objektem (vpravo dole) Výstup detekce duplicitních částí obrazu
original image same image with two figures deleted (bottom right) Detection of copy-move

Detection of noise inconsistency scans the image area and detects inconsistency in image noise characteristics, which hints that a given part of the image has been copied from another photo.

Původní autentický obrázek Obrázek s přidaným objektem Výstup detekce šumově konzistentních částí obrazu
authentic image Image with an inserted object that is "camouflaged" with noise Output of the detection of noise inconsistency

Chromatic aberration inconsistency is based on detecting chromatic aberrations (color blur) that are result of varying diffraction of different colors in camera optics. The inconsistency in aberration is another indication that the photo has been manipulated, such as cut/paste parts from other photos.

Original picture Modified picture Authentic chromatic aberration Irregular chromatic aberration The evaluation

 

Image and video restoration

This module is used to improve the visual quality and readability of digital images and videos.

Denoising removes noise from the input image or video.

 

Super-resolution estimates an image of higher resolution from several input images or video frames (according to the settings). For example, it can increase legibility of a text (such as license plates).

Jeden ze vstupních snímků s nízkým rozlišením Výsledek zvětšení rozlišení
One of the input image of low resolution Result of the SR algorithm

 

JPEG artifact removal eliminates visual defects that are apparent if the picture is saved as JPEG with low quality.

The input image (picture detail) Output of the filtration