Cloud-scale Image Compression Through Content Deduplication

David Perra and Jan Frahm

In Proceedings British Machine Vision Conference 2014


As more images are uploaded to the cloud, it is imperative that pixel-wise redundancy between images be leveraged to minimize each photo's memory footprint. We present an efficient cloud-scale digital image compression scheme for use in cloud image storage systems. Unlike current state-of-the-art systems, our image compression technique takes full advantage of redundant image data in the cloud by independently compressing each newly uploaded image with its GIST nearest neighbor taken from a representative set of uncompressed images. We leverage state-of-the-art video compression techniques in order to efficiently reuse image content which is already stored server-side. Our novel scheme scales to accommodate large datasets and is highly parallelizable for cloud computing. Experimental results demonstrate that our algorithm produces competitive image compression rates while reducing the computational effort by at least an order of magnitude, and providing scalability to accommodate the true scale of large-scale image compression.


Poster Session


Extended Abstract (PDF, 1 page, 213K)
Paper (PDF, 12 pages, 1.1M)
Bibtex File


David Perra, and Jan Frahm. Cloud-scale Image Compression Through Content Deduplication. Proceedings of the British Machine Vision Conference. BMVA Press, September 2014.


	title = {Cloud-scale Image Compression Through Content Deduplication},
	author = {Perra, David and Frahm, Jan},
	year = {2014},
	booktitle = {Proceedings of the British Machine Vision Conference},
	publisher = {BMVA Press},
	editors = {Valstar, Michel and French, Andrew and Pridmore, Tony}
	doi = { }