Digital Image Compression Using Bit Plane Slicing Method Póth Miklós Fürstner Igor Subotica Tech
Data compression Lossless - all original data can be recovered when the file is uncompressed. The signal is perfectly reconstructed from the available samples. (ZIP, GIF, PNG) Lossy – reduces a file by eliminating certain information. Permits reconstruction only of an approximation of the original data. (JPEG, MPEG, MP3) SiP 2017
Test images Cameraman Lena Clock MRI Mandril Einstein SiP 2017
2D-1D transformation Prior to compression, it is needed to transform the 2D image data to 1D data. Horizontal, vertical, Hilbert, …
Image scanning Horizontal Perimeter Zig-zag Z-curve Hilbert
Hilbert curve Fractal curve, self similar Hausdorff-Besicovich dimension 2 Capable of collecting all image pixels Applicable only to square images Image side must be a power of 2
Hilbert curve scanning Hilbert curve always takes adjacent pixel
Hilbert curve scanning Original image Cameraman Cameraman after Hilbert scanning
Differential encoding Run-length coding
Bit planes
Bit planes
Bit planes of Cameraman image MSB carries the contours of the image, LSB reminds of noise
Bit planes Question: How many bit-planes can be ignored? How will it effect the image?
Bit planes Question: How many bit-planes can be ignored? How will it effect the image?
Bit planes Question: How many bit-planes can be ignored? How will it effect the image?
Bit planes
Bit planes
Savings By ignoring 3 bit planes we save 3/8 = 37.5% of total image space.
Peak signal to noise ratio - PSNR ratio between the maximum possible power of a signal and the power of corrupting noise The signal in this case is the original data, and the noise is the error introduced by compression
Peak signal to noise ratio - PSNR It has been experimentally discovered that no disturbing visual artifacts are seen if PSNR>25 dB Quality is also based on image content, for some images it is 30 dB
Thank you for your attention. Questions?