Focuspath-uoft

Non-Reference Focus Quality Assessment (NR-FQA) metric for

• Synthetic blur imaging - Synthetic-MaxPol
• Natural blur imaging - HVS-MaxPol
• Out-of-focus in digital pathology - FQPath

Objective

One of the objectives of the project is to build a non-reference image sharpness assessment metric. The design of this metric is based on the notion of Human Visual System (HVS). We synthesized a Finite Impulse Response (FIR) kernel to mimic HVS response. We use this metric to score the blur level of the image, called HVS-MaxPol. We provide two versions of such metric. HVS-MaxPol-1 employs a single kernel yielding the best performance and HVS-MaxPol-2 employs two kernels. This metric is the state-of-the-art, which meets the accuracy and fast implementation at the same time. It is also applicable for both natural and synthetic images for focus quality assessment (FQA).

How to Cite

When using MaxPol for synthetic blur assessment, please cite:

M. S. Hosseini and K. N. Plataniotis, "Image Sharpness Metric Based on Maxpol Convolution Kernels," 2018 25th IEEE International Conference on Image Processing (ICIP), Athens, Greece, 2018. [Lecture Slides]

When using FocusPath-UofT for blur assessment, please cite:

Mahdi S. Hosseini, Yueyang Zhang and Konstantinos N. Plataniotis, “Focus Quality Metric Based on Visual Sensitivity,” submitted to IEEE Transactions on Image Processing, July 2018.

When using HVS-MaxPol for natural blur assessment, please cite:

Mahdi S. Hosseini, Yueyang Zhang and Konstantinos N. Plataniotis, “Focus Quality Metric Based on Visual Sensitivity,” submitted to IEEE Transactions on Image Processing, July 2018.

When using FQPath for digital pathology out-of-focus assessment, please cite:

Mahdi S. Hosseini, Yueyang Zhang, Lyndon Chan and Konstantinos N. Plataniotis, "Focus Quality Assessment of High-Throughput Whole Slide Imaging in Digital Pathology," submitted to IEEE Transactions on Medical Imaging, 2018.