eduzhai > Applied Sciences > Engineering >

PIPAL a Large-Scale Image Quality Assessment Dataset for Perceptual Image Restoration

  • king
  • (0) Download
  • 20210505
  • Save

... pages left unread,continue reading

Document pages: 18 pages

Abstract: Image quality assessment (IQA) is the key factor for the fast development ofimage restoration (IR) algorithms. The most recent IR methods based onGenerative Adversarial Networks (GANs) have achieved significant improvement invisual performance, but also presented great challenges for quantitativeevaluation. Notably, we observe an increasing inconsistency between perceptualquality and the evaluation results. Then we raise two questions: (1) Canexisting IQA methods objectively evaluate recent IR algorithms? (2) When focuson beating current benchmarks, are we getting better IR algorithms? To answerthese questions and promote the development of IQA methods, we contribute alarge-scale IQA dataset, called Perceptual Image Processing Algorithms (PIPAL)dataset. Especially, this dataset includes the results of GAN-based methods,which are missing in previous datasets. We collect more than 1.13 million humanjudgments to assign subjective scores for PIPAL images using the more reliable "Elo system ". Based on PIPAL, we present new benchmarks for both IQA andsuper-resolution methods. Our results indicate that existing IQA methods cannotfairly evaluate GAN-based IR algorithms. While using appropriate evaluationmethods is important, IQA methods should also be updated along with thedevelopment of IR algorithms. At last, we improve the performance of IQAnetworks on GAN-based distortions by introducing anti-aliasing pooling.Experiments show the effectiveness of the proposed method.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×