Behind the pixels ZSOLT OCSKAI DENNIS GABOR COLLEGE

Behind the pixels ZSOLT OCSKAI - DENNIS GABOR COLLEGE 21 -04 -2015

A few common things about raster images Their data are stored in pixels The pixels contain channels Each channel contains 8 or 16 bits of information The number of channels is determined by the used colour space: RGB – RED GREEN BLUE CMYK – CYAN MAGENTA YELLOW BLACK Palette – Indexed stored Colours (GIF) Etc…

A few common things about the raster images R: 90 G: 84 B: 90 Photo by Rita Szabó

Shannon entropy for digital images If we assume pixels are independent from each other, then we fit to use Claude E. Shannon’s entropy formula to process a digital image. possible range of values for the Entropy: Where

Shannon entropy for digital images Images properties: JPG 24 bits/pixels RGB HEIGHT 1000 960 WIDTH 643 667 ENTROPY 12, 453 Photo by Rita Szabó 15, 0286 Photo by Péter Vas

Differences between the image formats Images properties: FORMAT JPG TIFF 24 bits/pix. 48 bits/pix. RGB WIDTH HEIGH T 10000 ENTROPY 24, 7204 14, 8282

TAGGED IMAGE FILE FORMAT The TIFF is capable of: Carrying multiple images Storing 1, 2, 3, 4, etc channels/pixel Storing 1, 2, 3, 4, etc… bits/channel Multiple options to order the channels, bytes and bits Storing Layers For 4 GB physical storage boundary

TIFF – amount of channels for 1 pixel RGBA RGB hyperspectral multispectral Capable of carrying more than 359 channels

The 2 main byte - channel orders for 3 pixels in 1 row Per Interleaved Channel I II II II I I I

TIFF - head The TIFF is owned by the Adobe Inc. and it’s just partly opened format. The navigation map for the data structure can be found in the TAGs. All TAGs have unique address Some TAGs are forbidden to use / write without Adobe’s approval There are several un-official TIFF formats GEOTIFF BIGTIFF

The entropy compute software for TIFF images For the properly TIFF data reading we used the Lib. TIFF open source libraries The libraries are available for C++ and. NET environments too and mostly work in the same way We created several applications for more efficient With the programs we can: work - determine the entropy for the whole image - determine the entropy for a single channel The programs are available for x 64 systems: Images can be: - Windows - 8, 16 bits/channels - Linux - 1, 2, 3, 4, etc. Channels/pixel - OS X

The software I Visual Basic. NET Console and GUI For. NET framework with Lib. Tiff. NET C++ console and Java GUI For Windows, Linux and OS X enviroments with Lib. Tiff

The software II For JPG, PNG and BMP images C# for. NET 4. 0 framework Variants for x 86 and x 64 systems For educational purpose

Some unusable results White. tiff Number of pixels: 1000000 Number of unique colours: 1 Entropy value: 0 Test_8 bit_grey. tiff Number of pixels: 2560000 Number of unique colours: 256 Entropy value : 8 Black. tiff Number of pixels: 1000000 Number of unique colours: 1 Entropy value : 0 Test_24 bit_rgb. tiff Number of pixels: 2560000 Number of unique colours: 256 Entropy value : 8

Where it has been used The research project Educational Applications of Entropy and Fractal Structure Analysis in Impact Assessment of Environmental Pollutions (Journal of Applied Multimedia 2. /VIII. /2013) It was an agricultural project to analyse the possibilites of the pollution detections on images. The entropy showed no significant results The SFD (Spectral Fractal Dimension) was capable of detecting the different pollutions in the images

Important links „Digitális Fotósuli” - GDF Workshop SFD Information Technology Ltd. - Digital Image processing web page Adobe Systems Inc. - Offical TIFF web page Lib. Tiff project - TIFF handler packages for C++ (remotesensing. org) Lib. Tiff. NET project - TIFF handler packages for. NET Framework (Bit Miracle) Research project – researchgate. net