Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

High Resolution Image Quality Database

Huang Huang, Qiang Wan, Jari Korhonen

TL;DR

The paper addresses the lack of high-resolution BIQA data by introducing HRIQ, a database of 1120 natural images with MOS obtained from a controlled lab study involving 175 participants. It evaluates BIQA methods across multiple resolutions, showing that models trained on low-resolution data do not transfer well to high-resolution content, while deep learning approaches, particularly a new patch-based HR-BIQA, perform best on full-resolution images. The work demonstrates the necessity of high-resolution datasets and specialized models for accurate MOS prediction on high-res displays. The dataset and code are publicly available, enabling further development of high-resolution BIQA methods for applications such as 4K displays and professional imaging workflows.

Abstract

With technology for digital photography and high resolution displays rapidly evolving and gaining popularity, there is a growing demand for blind image quality assessment (BIQA) models for high resolution images. Unfortunately, the publicly available large scale image quality databases used for training BIQA models contain mostly low or general resolution images. Since image resizing affects image quality, we assume that the accuracy of BIQA models trained on low resolution images would not be optimal for high resolution images. Therefore, we created a new high resolution image quality database (HRIQ), consisting of 1120 images with resolution of 2880x2160 pixels. We conducted a subjective study to collect the subjective quality ratings for HRIQ in a controlled laboratory setting, resulting in accurate MOS at high resolution. To demonstrate the importance of a high resolution image quality database for training BIQA models to predict mean opinion scores (MOS) of high resolution images accurately, we trained and tested several traditional and deep learning based BIQA methods on different resolution versions of our database. The database is publicly available in https://github.com/jarikorhonen/hriq.

High Resolution Image Quality Database

TL;DR

The paper addresses the lack of high-resolution BIQA data by introducing HRIQ, a database of 1120 natural images with MOS obtained from a controlled lab study involving 175 participants. It evaluates BIQA methods across multiple resolutions, showing that models trained on low-resolution data do not transfer well to high-resolution content, while deep learning approaches, particularly a new patch-based HR-BIQA, perform best on full-resolution images. The work demonstrates the necessity of high-resolution datasets and specialized models for accurate MOS prediction on high-res displays. The dataset and code are publicly available, enabling further development of high-resolution BIQA methods for applications such as 4K displays and professional imaging workflows.

Abstract

With technology for digital photography and high resolution displays rapidly evolving and gaining popularity, there is a growing demand for blind image quality assessment (BIQA) models for high resolution images. Unfortunately, the publicly available large scale image quality databases used for training BIQA models contain mostly low or general resolution images. Since image resizing affects image quality, we assume that the accuracy of BIQA models trained on low resolution images would not be optimal for high resolution images. Therefore, we created a new high resolution image quality database (HRIQ), consisting of 1120 images with resolution of 2880x2160 pixels. We conducted a subjective study to collect the subjective quality ratings for HRIQ in a controlled laboratory setting, resulting in accurate MOS at high resolution. To demonstrate the importance of a high resolution image quality database for training BIQA models to predict mean opinion scores (MOS) of high resolution images accurately, we trained and tested several traditional and deep learning based BIQA methods on different resolution versions of our database. The database is publicly available in https://github.com/jarikorhonen/hriq.
Paper Structure (7 sections, 2 figures, 1 table)

This paper contains 7 sections, 2 figures, 1 table.

Table of Contents

  1. Introduction
  2. Database Creation
  3. Content Collection
  4. Subjective Quality Assessment Study
  5. Subjective Data Analysis
  6. Evaluating BIQA methods on HRIQ
  7. conclusions

Figures (2)

  • Figure 1: Comparison of the visual effects of blur at high and low resolutions. Figure (a) shows a patch of $512\times384$ pixels cropped from the original image of $2880\times2160$ pixels, and Figures (b) shows the corresponding full image resized to $512\times384$ pixels.
  • Figure 2: Subjective rating interface. The $2880\times2160$ resolution image displayed on a $3840\times2160$ monitor takes most of the screen. The user interface for rating is on the right side of the screen.