Visual Quality Assessment Databases, MMSP Konstanz
  • Home
  • VQA Databases
    • KonViD-150k Database
    • KonViD-1k Database
    • KonIQ-10k Database
    • KonIQ++ Database
    • KADID-10k Database
    • IQA-Experts-300
    • KonPatch-30k Database
    • KoSMo-1k Database
    • StudyMB 2.0 Database
    • Picturewise JND Data
    • KonJND-1k database
    • KonFiG-IQA Database
    • GFIQA-20k Database
  • About
  • Contact

KonPatch-30k IQA Database

Individually Annotated Image Patches

Image quality assessment (IQA) has been studied almost exclusively as a global image property. It is common practice for IQA databases and metrics to quantify this abstract concept with a single score per image. In an attempt to extend the notion of quality to spatially restricted sub-regions of images, we designed a novel database of individually quality-annotated image patches.

We randomly sampled 500 images from KonIQ-10k as a data source and selected 64 patches of 64x64 pixels from each source image. Each patch was annotated in two separate experiments. We conducted a lab study where we collected a single vote per patch in an controlled environment. The consideration to be made was whether the presented patch looks indicative of being sampled from a high quality image. This experiment was repeated in a crowd study with 10 votes per patch, resulting in a dataset with 32,000 patches and a total of 352,000 binary votes.


Cite us

KonPatch-30k is freely available to the research community. If you use our database in your research, please cite the following reference:
@inproceedings{wiedemann2018disregarding,
title={Disregarding the Big Picture:
Towards Local Image Quality Assessment},
author={Wiedemann, Oliver and Hosu, 
Vlad and Lin, Hanhe and Saupe, Dietmar},
year={2018},
organization={IEEE},
booktitle = {10th International Conference on 
Quality of Multimedia Experience(QoMEX)},
url = {http://database.mmsp-kn.de}
}
​
Picture

Downloads

Images (.zip)
180MB download
SCORES (.CSV)
3 Mb download

The scores file

  • id: Identifier for each patch, unique within this database.
  • image: Corresponds to a source file from KonIQ-10k.
  • x, y: Pixel coordinates where the patch was sampled within the source image.
  • koniq10k_mos: Mean opinion score of the source image (copied from KonIQ-10k).
  • labvote: Binary lab vote whether the patch is indicative of high quality or not.
  • crowd_upvotes: Number of crowdworkers that rated this patch as rather high quality.
  • crowd_downvotes: Number of crowdworkers that rated this patch as rather low quality.

Examples

Rather low quality

Rather high quality

Imprint | Privacy policy | About | Contact
Copyright "VQA Group at Universität Konstanz" © 2019-2022
  • Home
  • VQA Databases
    • KonViD-150k Database
    • KonViD-1k Database
    • KonIQ-10k Database
    • KonIQ++ Database
    • KADID-10k Database
    • IQA-Experts-300
    • KonPatch-30k Database
    • KoSMo-1k Database
    • StudyMB 2.0 Database
    • Picturewise JND Data
    • KonJND-1k database
    • KonFiG-IQA Database
    • GFIQA-20k Database
  • About
  • Contact