THE IMPLEMENTATION OF DEPERSONALIZATION ALGORITHM OF DIGITAL IMAGES
There needs to ensure confidentiality
of digital images created by using photo and video cameras in various spheres of
There exist the various identifying
signs of digital optoelectronic devices (DOED) . These signs are depended on
their individual characteristics. Besides of this the identification information
is saved on the images during all period of post-processing. That is why the
photo technical examination of the digital images as usual is performed by two
directions : research on digital images and research on photographic
The task of the research
is to develop the algorithm of depersonalization of digital images,
by removing the identification features of DOED.
Identification features of DOED
The development of methods to ensure
the author confidentiality of digital images are one of the urgent tasks of
In most cases the decision on the
identification of DOED, used to create the digital
image, is taken providing that the digital image under examination is suitable
for being identified.
DOED identification data that can be
obtained during the examination of the
· setting of shooting conditions;
· determination of the image creation time;
· restoration of the original image;
· the type of DOED used to produce the image;
· subsequent digital image processing (in the
graphic editors and media converters);
· assessment of the changes in the graphic image
· the type of the software used for image
General studies for documenting DOED
characteristics reflected in the created digital image include frame sizes and
The digital image in JPEG format, the
creation algorithm of which is based on the discrete cosine transform (DCT), is
applied to the matrix of the original image to produce a new matrix of
coefficients (the digital image in a new format) . Therefore, the DCT
coefficients of the resulting matrix of the digital image can be attributed to
the group features of DOED. They are specific to its model or a series of models
. DCT matrix of the digital images and complete information about it are
called EXIF data . This data includes additional information (metadata) about
the digital image.
In general, the EXIF data include:
· name of DOED model;
· DOED orientation (vertical / horizontal – for
devices with built-in accelerometer);
· address of the shooting locations – the
position on the earth's surface;
· date and time of shooting;
· size of the digital image;
· digital image resolution;
· depth of color in bits;
· type of white balance;
· focal distance;
· equivalent focal distance – the common
characteristic of the optical system and the light-sensitive element, which
gives information about the viewing angle of the system;
· flash use;
· ISO – the sensitivity of DOED;
· the sensitivity of the sensor (the matrix),
which is set while shooting;
· the software, in which the digital image was
made (if the image is in JPG format, there will be indicated the DOED software
which processed the image; if the image is in RAW format there will be indicated
the software which exported it to JPG format, for example, Adobe Photoshop
· exposure time in seconds while the digital
image was photographed. It’s the interval of time during which the optical
system transmits the image to the photosensitive matrix;
· exposure time compensation;
· information on the right of possession.
media files can include the following EXIF data: video data; audio data;
flash-content (SWF format); Categories – contains information based on
Analysis of digital image EXIF data
The analysis of the digital image
EXIF data can be carried out using JPEGSnoop software which supports the
following image formats: JPG, THM, AVI, DNG, CRW, CR2, NEF, ORF, PEF, RAW, MOV
and PDF, Photoshop files . These EXIF data are exported to TXT file. One of
the important functions of the JPEGsnoop software is the presence of the
internal database that compares the tested digital image with the large number
of compression signatures. It’s digital signature, with which the resulting
digital image is marked during the compression by different algorithms. The
digital signature is placed in the noise components of the image. This enables
to uniquely determine which DOED was applied to form the digital image. The
JPEGsnoop software retrieves the following information from the digital image:
· quantization of the matrixes;
· the color of sub-sampling;
· the quality of JPEG container;
· the settings of JPEG container resolution;
· Huffman tables. Basing on this table, the
coding tree is formed. (The classic Huffman algorithm that receives the
frequency table of symbol occurrence in a message. This algorithm is used for
compression of textual and graphical information);
· Makernotes. Every digital optoelectronic
device stores the EXIF information in the extended section. This information is
specific to its manufacturer and is contained in the so-called Maker Note
· RGB histogram – uses all three-color channels
and describes the brightness distribution in a single channel and shows the loss
in a separate chrominance channel, but it does not show whether the losses are
in one or all channels. Color histograms amplify this effect and clearly
demonstrate the presence of losses;
· markers JPEG (JFIF) – JPEG files contain a
sequence of markers. Each of them begins with the 0xFF byte, indicating the
beginning of the marker, and with the ID-byte. This JPEG-file structure allows
you to quickly find the marker with the necessary data (for example, the length
of string, the number of strings and the number of color components of the
· VLC decoder type (VLC decoding – method of
adaptive coding of variable length (VLC) with efficient memory and low
complexity for data of various applications, such as coding of digital video
data, image data, audio or speech data);
· determination of the quality parameters used
in Photoshop software;
· retrieving of embedded images in Adobe PDF
Individual signs of DOED
The following properties and features
for hardware DOED can be defined and used for DOED identification:
1. For digital photo, video,
· lens and bayonet mounting system (forms medium
steady signs). Bayonet – the kind of connection to fix the optical system (lens)
to digital optoelectronic devices. Bayonet is not only mechanical but also
electronic interface. It connects microprocessors of the lens and digital
optoelectronic device via electrical contacts;
· blur – irreversible operation, when the
digital image or its parts are redistributed according to some law (forms
medium- and high-resistant signs).
The strength of the blur is effected by these physical parameters.
Geometric lens aperture (F): the
smaller the number of F, the thinner the depth of the field and the stronger the
blur of the foreground and background images.
The focal distance of the lens: the
larger the focal length is, the stronger the background image is blurred.
The focus distance to the subject
(the distance between the camera and the subject shooted): the smaller the focus
distance, the stronger the blur of the background image.
The distance between the subject and
the background: the farther the background of the subject, the stronger it is
Optical scheme: it has a greater
effect on the blur type.
The size of light-sensitive matrix
module (photosensitive matrix which presents an integrated circuit, consisting
of millions of pixel cells (photodiodes): photodiodes are able to convert light
energy into electric charge stream. It can be read and amplified by an
analog-digital convertor and is converted into a predetermined bit binary code.
Then this code goes to the digital processor of the optoelectronic device for
subsequent processing). The larger the size is, the larger the angle of the view
is and the closer you need to come to the subject. Therefore,
DOEU blurs the background image more strongly.
Special nozzles and filters on the
The photo sensor: it forms stable
2. For digital scanners:
of light-sensitive elements of the scanner line;
· deviation of the scanner carriage movement
· uneven illumination and uneven pressing to the
glass of scanned original, etc;
· in addition to DOED features mentioned above,
their individual features are largely determined by the applicable
built-processing algorithms of the created digital images:
1. Image reconstruction algorithms from a mosaic photo sensor structure – is
based on measuring only one color component at each point of photo sensor, and
the missing components are calculated basing on the neighboring data points.
This technology creates a photo sensor which measures three colors at each pixel
2. Improving of the contour sharpness (the algorithm reproduces the traditional
technique of the film mask. This mask is used to increase the sharpness of edges
in the image. The algorithm corrects blurring of the digital image that appears
in the result of scanning, printing or interpolation) and noise reduction (the
algorithm of suppression of all detected noises which have been generated in the
digital image before its being recorded).
The post-processing algorithms of
creating digital images by the digital photo, video and web-cameras can be both
disengaged and engaged.
In digital scanners,
a resulting image can pass two-level process – in the scanner itself with the
help of calibration curves, the suppression of dust on the scanned copy and
scanner glass, and at the driver level, where the subjective improvement of the
quality of the digital image is carried out.
(regardless of the format produced by the digital image) has its own
frequency-contrast characteristic, which is type of amplitude-frequency
characteristic (AFC) of the optical system and a photosensor. The
amplitude-frequency characteristic is not constant in different parts of the
digital image, and its correct comparison is best carried out by the pictures of
the radial test-object (the test-object is used for the quantitative
determination of the resolution and for the modulation transfer of the lens and
supersensitive matrix), but it is not always feasible on practice. Therefore, to
the individual features of DOED can belong its pattern of the photo-response
non-uniformity; the presence and location of contamination on its photo sensor
(photo matrix), which is transferable to the digital image (Fig. 1). This
feature is inherent to the mirror type of DOED, as, while replacing the lens,
the photo sensor inevitably gets covered in dust.
Fig. 1. The presence of impurities on the matrix of the photosensor DOED (arrows
indicate the presence of dust impregnations).
In addition, there are cases when the
surface of the optical system in the non-separable photo, video, web-cameras and
scanners is contaminated (Fig. 2).
Fig. 2. Characteristic surface contamination of the optical system.
The use of visible and invisible
digital watermark is a common method of protection against unauthorized copying
of digital content. This also makes it possible to uniquely identify the DOED in
the digital image that it creates (Fig. 3) .
Implementation of the digital
watermark is carried out by using the following criteria:
· digital watermark is introduced into the
digital image by the special algorithm, which does not allow to determine the
presence of the digital watermark in the image;
· the presence of the secret key;
· the possibility of proving the existence of a
digital watermark to a third party without disclosing the secret key;
· digital watermark maintains all types of image
distortions or the maximum possible number of images, except for those that make
it practically useless for use;
the developed digital watermark models are
resistant to any changes in the digital image.
The results of detecting and extracting the digital watermark from the
modified image: a) the map of the detected modified blocks (1,2,3a, 3b); b) the
modified blocks (marked in black); c) the extracted digital watermark (white
color - damaged blocks, and light gray – restored digital watermark – IPSI
Identification of DOED by the digital noise
is a defect of the digital image, made by the DOED photo sensor. The digital
noise shows itself in the form of randomly spaced points of different brightness
and color. It is especially noticeable on plain surfaces – the sky, the skin,
and the shadow areas. The digital noise gives the digital image unnatural
appearance – it seems to be "sprinkled" with sand . The digital noise is
typical of each matrix of the optoelectronic digital device. It becomes clearly
visible when increasing the sensitivity (ISO 400, 800, 1600, etc.).
The digital noise is usually
associated with an electrical error of the photo
sensor (matrix). This phenomenon arises from the individualities of the
refraction of light – there appear
multi-colored pixels on the matrix. The more pixels are located on the matrix,
the smaller their size is. Increasing the amount of the pixels on the matrix
results in increasing of their sensitivity. It stimulates the noise increase. It
should also be noted that while increasing the sensitivity (ISO), the
temperature of the photo sensor becomes also increased, which contributes to the
developing of the noise. The noise level depends on the technical
characteristics of the photo sensor and the duration (time) of the exposure. The
digital noise is divided into: permanent, incidental, luminance and chrominance.
Permanent digital noise equally appears on all digital images created by
DOED and is connected with the presence of "hot" and "broken" pixels of its
photo sensor. Hot pixels – occur in the form of colored pixels and depend on the
photo sensor temperature. During the work of the matrix the temperature of
pixels rises and the pixels, unsustainable to high temperature, start to "act
up" giving the signal which at times may be different from normal neighboring
pixels and to be brighter or darker than you need. Hot pixels are usually red,
blue or green. Broken pixels – appear regardless of the shooting mode and may be
bright or very dark, depending on the mode they lost their efficiency in. If the
broken pixel "floats" in the switched-on state, there will be bright pixels
(usually white), because it takes too much light. If the pixel "floats" in the
off position, then the pixel will be dark (almost black), because it is very
poorly reactive to light. In the place of the "broken" pixels there are always
bright or dark spots. "Hot" pixels appear as colored dots, positioned at the
same place from frame to frame during long exposures when the photo sensor is
extremely hot. For identification of the "hot" pixels it is necessary to cover
DOED with dustproof camera lens cap, to select the maximum ISO sensitivity
value, set the shutter speed of 30 seconds, turn off the built-in noise
reduction (if any) and make some control of digital images. Their visual
analysis shows the presence or absence of "hot" pixels.
Luminance digital noise is shown on the image in the form of small dark
spots (or points) and resembles a film grain. Film grain – grain size of
the image is increased due to the process of increasing the film sensitivity.
Large grains reduce the resolution of the film.
is shown on the image in the form of small spots (points) of a different color,
different from the color of the areas where this noise is shown (that is why it
is clearly visible). Chrominance digital noise is striking and unpleasant for
Depersonalization algorithm of digital images
Depersonalization of the digital
image means absence of identifying features that can help to uniquely identify
DOED. The sequence of actions is proposed and carried out by the authors for the
depersonalization of the digital image. It is presented in the form of the
algorithm in Fig. 4.
Fig.4. Depersonalization algorithm of digital images.
Creating a digital image with the lowest
possible level of digital noise. Among recommendations to reduce
primary noise generated in the digital image there are conventional steps to
reduce internal noise of the photo sensor, namely :
· decreasing of the ISO sensitivity;
· reducing of exposure;
· use of
a high aperture lens (if the aperture is opened wider, the exposure will be
· shooting under a bright light (or using the flash);
· using a
built-in noise suppression function;
· avoiding too long operating of DOED without its being off (it causes heating
of the photo sensor, especially of mirror cameras
in the focus mode of LCD display);
of digital images in RAW format. This is the format of the digital image
containing the raw data received from the photo sensor. Complete information
about the stored information signal is contained in such files. This
information has no precise specifications (standard). It is also sometimes
referred to as "raw" format.
The implementation of specialized software is carried out in
the proposed algorithm of depersonalization of the digital images following the
Removing of "broken" and "hot" pixels
If DOED creates the digital image in RAW format, the task of
removing the "broken" and "hot" pixels is greatly simplified by using
specialized software – Hot Pixel Eliminator (the
software can distinguish between "hot" and "broken" pixels from the bright glare
and light sources) , Pixel Fixer (the
software can automatically remove hot and broken pixels from the raw digital
If the specialized software is not able to remove the "broken" and "hot" pixels,
or if the generated digital image is saved in the JPEG-format, you can remove
them manually in Adobe Photoshop graphics editor by using Patch Tool.
Removing of digital noise
Universal algorithms of removing/suppressing the digital
noise have not been yet developed. The specialized software, that implements
these algorithms, cannot always distinguish fine details of the digital image
from the digital noise. Consequently, strong suppression of the digital noise
often leads to the partial loss of little bits and is shown in the form of
digital image blur. The specialized software used for the removal/ suppression
of the digital noise is to consider the following factors:
· the model of DOED;
for the photo sensor digital noise;
presence of little bits on the digital image.
The specialized software should be able to manually set the
modes of disposal/ suppression of the digital noise, as the perception of
digital images by the human eye is subjective. The practical use of this
software shows that, with the user's appropriate experience skills, better
results can be achieved in manual mode than in automatic one.
recommendations of the specialized software applications should include the
implementation of the procedure of the noise deletion / suppression before other
operations such as color correction, brightness / contrast, resize, etc. There
are a lot of specialized commercial software that can be used to suppress the
digital noise, such as Adobe Camera RAW, Adobe Light Room and others. The
specialized commercial software Movavi Photo Denoise was used by the authors to
test the removal operation of the digital noise (Fig. 5) .
The result of the practical implementation defects removal and
photosensor digital noise correction: a) the existence of "broken" pixel in
digital image; b) correction photosensor digital noise in digital image.
Removing of EXIF data
Removing of EXIF data helps to depersonalize DOED according
to its inserted metadata into the digital image during its creation, and to hide
information about specific software that the digital image has been processed by
All the specialized software used to perform operations with
the EXIF data can be divided into three groups, each consisting of specific
· metadata reviewing;
· selected metadata tags editing;
· complete removing of metadata.
It should be noted that the operation of removing the EXIF
data is implemented by the same principle as editing.
The most effective
universal means for this operation is ExifTool software.
This tool is available
for all platforms and recognizes additional tags (EXIF chunks) of the most DOED
and software used for post-processing of the digital images .
The algorithm of depersonalization of digital images was
developed during researching of various characteristics. These characteristics
uniquely identify of DOED.
The practical usage of the developed
algorithm allows to depersonalize of the digital image by removing the
identification features of DOED.
The algorithm can be useful for those who perform their
professional duties for information security and others for personal security.
Ustanovlenie avtorskih prav po neodnorodnostjam cifrovyh obrazov /
Rubljov D.P., Fjodorov V.M., Chumachenko A.B., Makarevich O.B. // Izvestija
Juzhnogo federal'nogo universiteta. Tehnicheskie nauki. – 2008. – № 8. – T. 85.
– S. 141–147.
Fototehnicheskaja jekspertiza. URL: http://www.expertsud.ru/content/view/173/.
(Cifrovoj shum na fotografii. URL: http://www.colorpilot.ru/reduce_noise.html.
(Data of reference: 25.02.2017).
Jeffektivnye reshenija dlja zashhity ot moshennichestva v cifrovom mire.
URL: http://www.smtdp.com. (Data of reference: 25.02.2017).
Identifikacija cifrovogo fotoapparata. URL: https://www.pgpru.com/forum/anonimnostjvinternet/identifikacijacifrovogofotoapparata?show_comments=1&p=2#Comment69027.
(Data of reference: 25.02.2017).
Chto takoe Exif cifrovogo fotosnimka. URL: https://profotovideo.ru/polezno-znat/chto-takoe-exif-tsifrovogo-fotosnimka.
(Data of reference: 25.02.2017).
JPEGsnoop 1.7.5 – JPEG File Decoding Utility. URL: http://www.impulseadventure.com/photo/jpeg-snoop.html.
(Дата of reference: 25.02.2017).
Glumov Nikolaj Ivanovich, Mitekin Vitalij Anatol'evich Algoritm
vstraivaniya poluhrupkih cifrovyh vodyanyh znakov dlya zadach autentifikacii
izobrazhenij i skrytoj peredachi informacii // KO. – 2011. – №2. – S. 262–267.
Cifrovoj shum. Cifrovoj shum – chto jeto takoe? Sposoby bor'by s cifrovym
shumom. URL: http://lepser.ru/teoriya-fotografii/tsifrovoy-shum-chto-eto-takoe-sposobyi-borbyi-s-tsifrovyim-shumom.html
(Data of reference: 25.02.2017).
Cifrovoj shum na fotografii. URL: http://www.colorpilot.ru/reducenoise.html.
(Data of reference: 25.02.2017).
HotPixels Eliminator. URL: http://www.mediachance.com/digicam/hotpixels.htm
(Date of reference: 25.02.2017).
Pixel Fixer. URL: http://pixelfixer.org (Date of reference: 25.02.2017).
Movavi Photo Denoise. URL: https://www.movavi.ru/noise-reduction-software/
(Date of reference: 25.02.2017).
Read, Write and Edit Meta Information! URL: http://www.sno.phy.queensu.ca/~phil/exiftool/
(Date of reference: 25.02.2017).
Source: Authors publication