Difference Between Full-reference, Half-reference and No-reference Image Quality Assessment

    Image quality assessment is to find out the distortion content and distortion degreebetween the fault image and the reference image.

    Image quality assessment(IQA) can be classified into three categories,full-reference(FR), half-reference(HR), and no-reference(NR), according towhether a perfect reference image is used.

   The following is a table to show thedifference between FR, HR and NR in three aspects.

 

full-reference(FR)

half-reference(HR)

no-reference(NR)

 

Concept

Requires full access  to the reference/original images in assessing the fault image.

Requires a reference image to participate partly in the assessment of the fault image.

Do not need the of the reference image.

 

Application

More applied in the occasion where the reference image could be easily got.

Such as comparison of the advantages of compression, fusion algorithms.

More applied in the occasion where the reference image is hard to get, but some extracted features of the reference image can be easily got.

Such as channel transmission, compression in special occasion.

More applied in the occasion where the reference image is not available, or the acquisition costs too much.

Such as the quality assessment of remote sensing images.

 

Feature

Low cost, fast.

The assessment result and the visual effect is not always consistent.

Higher efficiency, less cost of time and space than FR.

Narrower applications.

Wide range of applications, broad prospects of development.

Most difficult to study.