Skip to content

Image Vectorquantizer designed for Proseminar Datenkompression SS20 at RWTH Aachen University

Notifications You must be signed in to change notification settings

simonmichau/Image-Vectorquantizer

Repository files navigation

Image Vectorquantizer

designed for Proseminar Datenkompression SS20 at RWTH Aachen University

General info

IQCv2 takes a .jpg image and treats the individual RGB values of each Pixel in the image as a threedimensional vector. During encoding, a variation of the Linde-Buzo-Gray Algorithm is used to produce a codebook of a size that can be specified by the user. Supposing a naive image format, that stores each pixel color as three individual values between 0 and 256 (this would result in a codebook with 16.777.216 vectors and 24 bit for each pixel, if each value was stored in binary), this method of compression is capable of a significant saving of storage data. Actually, depending on the specific image, using the Linde-Buzo-Gray (LBG) Algorithm, as little as 128 codevectors can be enough to reconstruct an image that is basically indifferentiable to a human observer, as you can see below.

Original image 8 codevectors 64 codevectors 512 codevectors

The top left picture is the original one and the following are reconstructions from codebooks with 8, 64 and 512 codevectors. The encoding process will result in the files codebook.txt, where the binary values of the algorithmically determined codevectors are stored, and the file transmission.txt, where the binary index of the codevector that corresponds with each pixel is stored. During decoding, first the codebook is read, and then each value from the transmission.txt file is matched up with their respective index from the codebook, which is then used to reconstruct the image.

Complexity

Space Complexity File
equation for codebook.txt
equation for transmission.txt

where equation is the number of codevectors.

Runtime Complexity
equation for Encoding
equation for Decoding

Sources

The picture examples from this readme.md are outputs of this project for pictures licensed under Creative Commons (https://search.creativecommons.org/photos/9e65a3ae-8ea0-4637-8cb4-6d996624f96f)

About

Image Vectorquantizer designed for Proseminar Datenkompression SS20 at RWTH Aachen University

Topics

Resources

Stars

Watchers

Forks

Packages

No packages published

Languages