کنفرانس بین المللی فناوری اطلاعات و دانش

English

صفحه اصلی / شانزدهمین کنفرانس بین المللی فناوری اطلاعات و دانش

Scattering Wavelet-Based Image Quality Assessment Metric for Medical Images

نویسندگان :

Sina Omidvar¹ Jamshid Shanbehzadeh²

1- دانشگاه خوارزمی 2- دانشگاه خوارزمی

کلمات کلیدی :

image quality assessment،full-reference،medical imaging،scattering wavelet transform

چکیده :

In the medical domain, full-reference image quality assessment (FR-IQA) plays a vital role in clinical practice and directly affects diagnostic accuracy and efficiency. Conventional metrics like PSNR and SSIM are commonly used in general tasks, but often misalign with human perception of medical images, leading to possible hazards in clinical applications. To address this, we propose a novel FR-IQA metric based on the scattering wavelet transform (SWT) that extracts invariant and multi-scale features to capture complex textures and patterns. Our method computes the distance between extracted features from pristine and distorted images as the quality index. On the CSIQ database, our method achieves a high Spearman rank-order correlation coefficient (SROCC) of 0.94, rivaling top-tier metrics. In order to conduct a clinically relevant experiment, we also evaluate on a small dataset of 50 magnetic resonance (MR) images with real-world clinical distortions. Our method achieves a promising SROCC of 0.85, outperforming other metrics and showing potential for further research.