Image Sampling for Machine Vision

  • Jiashuai Cui
  • , Fan Li
  • , Liejun Wang

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

Abstract

Image sampling is one of the basic methods for image compression, which is efficient for image store, transmission, and applications. Existing sampling methods are designed for human-eye perception, which discard unconcerned information to decrease the amount of data considering the visual preference of human. However, these methods cannot adapt to the increasing machine vision tasks since there is a lot of redundant information for machine analysis to ensure the comfort of human eyes. In this paper, we propose an image sampling method for machine vision. We adopt a gray image to retain the main structural information of the image, and construct a concise color feature map based on the dominant channel of pixels to provide color information. Experiments on public datasets including COCO and ImageNet show that our sampling method can adapt to the characteristics of machine vision and greatly reduce the amount of data with little impact on the performance of mainstream computer vision algorithms.

Original languageEnglish
Title of host publicationArtificial Intelligence - Second CAAI International Conference, CICAI 2022, Revised Selected Papers
EditorsLu Fang, Daniel Povey, Guangtao Zhai, Tao Mei, Ruiping Wang
PublisherSpringer Science and Business Media Deutschland GmbH
Pages229-241
Number of pages13
ISBN (Print)9783031204968
DOIs
StatePublished - 2022
Event2nd CAAI International Conference on Artificial Intelligence, CAAI 2022 - Beijing, China
Duration: 27 Aug 202228 Aug 2022

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume13604 LNAI
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference2nd CAAI International Conference on Artificial Intelligence, CAAI 2022
Country/TerritoryChina
CityBeijing
Period27/08/2228/08/22

Keywords

  • Color feature
  • Image sampling
  • Machine vision

Fingerprint

Dive into the research topics of 'Image Sampling for Machine Vision'. Together they form a unique fingerprint.

Cite this