No-reference evaluation method for aerial image restoration quality based on joint learning

An aerial image and evaluation method technology, applied in image enhancement, image analysis, image data processing and other directions, can solve the problem of no reference evaluation of aerial image restoration quality, and achieve good prediction performance and good prediction effect.

CN111179224APending Publication Date: 2020-05-19中国人民解放军陆军炮兵防空兵学院
3 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Current Assignee / Owner
Publication Date
2020-05-19

Smart Images

  • Figure 1
    Figure 1
  • Figure 2
    Figure 2
  • Figure 3
    Figure 3
Patent Text Reader

Abstract

The invention discloses a no-reference evaluation method for aerial image restoration quality based on joint learning. The method comprises the following steps: firstly, learning distortion characteristics of a restored image by taking a convolutional neural network for layer-by-layer down-sampling as a main framework and taking an image classification task as a drive; and then regressing the distortion characteristics of the restored image to the real score of the image quality by using a joint loss function so as to realize approximation of the full-reference evaluation index SSIM of the image. According to the method, the problem of aerial image blind evaluation when the reference image cannot be obtained is solved, good prediction of image quality difference is considered, and the prediction performance of aerial image restoration quality is improved.
Need to check novelty before this filing date? Find Prior Art

Description

technical field

[0001] The invention relates to the technical field of image quality evaluation, in particular to a no-reference evaluation method for aerial image restoration quality based on joint learning. Background technique

[0002] Aerial images are an important source of geographic information. However, thick clouds and lens smudges will degrade the image quality and incomplete target information. When the spectral and temporal information is not enough to restore, people often use restoration methods to reconstruct the lost Information. The selection of restoration methods and the adjustment of parameters depend on the evaluation of restoration quality. Image quality evaluation is one of the basic problems in image processing, which provides evaluation indicators for image restoration, super-resolution reconstruction, defogging and rain removal. However, the current evaluation methods for image restoration quality are mainly full reference indicators such as SSIM,...

Examples

Embodiment 1

[0072]The experimental data in this embodiment is an aerial image of a certain region in Asia, which is cut into a total of 16,542 images of 256×256 without repetition, and divided into a training set and a test set at a ratio of 3:1. Three recent image restoration methods based on deep learning were selected [Yeh R A, Chen C, Yian Lim T, et al.Semantic image inpainting with deepgenerative models[C] / / Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.2017: 5485-5493][Pathak D,Krahenbuhl P,Donahue J,etal.Context encoders:Feature learning by inpainting[C] / / Proceedings of the IEEE conference on computer vision and pattern recognition.2016:2536-2544][Iizuka S,Simo -Serra E, Ishikawa H.Globally and locally consistent imagecompletion[J].ACM Transactions on Graphics(ToG),2017,36(4):107] Learning on the training set, these methods are based on deep learning (after extensive testing, When non-deep learning repair methods are used for aerial images, instabilit...