A parallel distributed processing algorithm for image feature extraction

Alexander Belousov, Joel Ratsaby

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

2 Scopus citations


We present a new parallel algorithm for image feature extraction. which uses a distance function based on the LZ-complexity of the string representation of the two images. An input image is represented by a feature vector whose components are the distance values between its parts (sub-images) and a set of prototypes. The algorithm is highly scalable and computes these values in parallel. It is implemented on a massively parallel graphics processing unit (GPU) with several thousands of cores which yields a three order of magnitude reduction in time for processing the images. Given a corpus of input images the algorithm produces labeled cases that can be used by any supervised or unsupervised learning algorithm to learn image classification or image clustering. A main advantage is the lack of need for any image processing or image analysis; the user only once defines image-features through a simple basic process of choosing a few small images that serve as prototypes. Results for several image classification problems are presented.

Original languageEnglish
Title of host publicationAdvances in Intelligent Data Analysis XIV - 14th International Symposium, IDA 2015, Proceedings
EditorsTijl De Bie, Matthijs van Leeuwen, Elisa Fromont
Number of pages11
StatePublished - 2015
Event14th International Symposium on Intelligent Data Analysis, IDA 2015 - Saint Etienne, France
Duration: 22 Oct 201524 Oct 2015

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349


Conference14th International Symposium on Intelligent Data Analysis, IDA 2015
CitySaint Etienne


Dive into the research topics of 'A parallel distributed processing algorithm for image feature extraction'. Together they form a unique fingerprint.

Cite this