Influence of Averaging Preprocessing on Image Analysis with a Markov Random Field Model
書誌事項
- 公開日
- 2018-02
- 資源種別
- journal article
- DOI
-
- 10.7566/jpsj.87.024802
- 公開者
- Tokyo : Physical Society of Japan
この論文をさがす
説明
This paper describes our investigations into the influence of averaging preprocessing on the performance of image analysis. Averaging preprocessing involves a trade-off: image averaging is often undertaken to reduce noise while the number of image data available for image analysis is decreased. We formulated a process of generating image data by using a Markov random field (MRF) model to achieve image analysis tasks such as image restoration and hyper-parameter estimation by a Bayesian approach. According to the notions of Bayesian inference, posterior distributions were analyzed to evaluate the influence of averaging. There are three main results. First, we found that the performance of image restoration with a predetermined value for hyper-parameters is invariant regardless of whether averaging is conducted. We then found that the performance of hyper-parameter estimation deteriorates due to averaging. Our analysis of the negative logarithm of the posterior probability, which is called the free energy b...
収録刊行物
-
- Journal of the Physical Society of Japan
-
Journal of the Physical Society of Japan 87 (2), 024802-, 2018-02
Tokyo : Physical Society of Japan
- Tweet
詳細情報 詳細情報について
-
- CRID
- 1521699230412352640
-
- NII論文ID
- 40021463387
- 210000134710
-
- NII書誌ID
- AA00704814
-
- ISSN
- 00319015
- 13474073
-
- NDL書誌ID
- 028822962
-
- 本文言語コード
- en
-
- 資料種別
- journal article
-
- NDL 雑誌分類
-
- ZM35(科学技術--物理学)
-
- データソース種別
-
- NDLサーチ
- Crossref
- CiNii Articles
- KAKEN
- OpenAIRE
