k-means Image Compression

Abstract of the work

The idea is to used Unsupervised learning technique, k-means clustering to reduce the size of the image. One utilises k-means clustering technique to cluster various colour which represent the image, or simplistically, forms the basis for the image class. For an input image, k-means clustering results in just 16 colour representation of the image. The image on the right shows the lossy-compression of the image with just 16 colours. Obviously one can change the number for the compressed representation but one has to take good care for the quality of the representation of the image. Initially, we test k-means on random data followed by image.

Support or Contact

In case you're interested to know more about it, contact at priyankt68@gmail.com or send a message to @priyankt68