Improving assaulted medical image quality using improved adaptive filtering network
by Namita D. Pulgam; Subhash K. Shinde
International Journal of Intelligent Systems Technologies and Applications (IJISTA), Vol. 22, No. 1, 2024

Abstract: Regular static convolutions work well for low-frequency information processing but fall short for high-frequency information processing. Dynamic convolution is the recent method that has spatial anisotropy and content-adaptiveness, enabling it to restore complicated and sensitive high-frequency information. The proposed method makes use of dynamic convolution to enhance the learning of multi-scale and high-frequency features. To accomplish this, two blocks - the dynamic convolution block (DCB) and the multi-scale dynamic convolution block (MDCB) are introduced. Dynamic convolution is used by the DCB to improve high-frequency information, whereas skip connections are used to protect low-frequency information. To efficiently extract multi-scale features, the MDCB uses shared adaptive dynamic kernels of increasing size along with dynamic convolution. The proposed multi-dimension feature integration mechanism is used to produce accurate and contextually enriched feature representations. For successful denoising, an improved adaptive dynamic filtering network is useful.

Online publication date: Mon, 05-Feb-2024

The full text of this article is only available to individual subscribers or to users at subscribing institutions.

 
Existing subscribers:
Go to Inderscience Online Journals to access the Full Text of this article.

Pay per view:
If you are not a subscriber and you just want to read the full contents of this article, buy online access here.

Complimentary Subscribers, Editors or Members of the Editorial Board of the International Journal of Intelligent Systems Technologies and Applications (IJISTA):
Login with your Inderscience username and password:

Username:        Password:          Forgotten your password?

Want to subscribe?
A subscription gives you complete access to all articles in the current issue, as well as to all articles in the previous three years (where applicable). See our Orders page to subscribe.

If you still need assistance, please email subs@inderscience.com