Quality assessment metrics for edge detection and edge-aware filtering: A tutorial review

Abstract

Abstract-The quality assessment of edges in an image is an important topic as it helps to benchmark the performance of edge detectors, and edge-aware filters that are used in a wide range of image processing tasks. The most popular image quality metrics such as Mean squared error (MSE), Peak signal-tonoise ratio (PSNR) and Structural similarity (SSIM) metrics for assessing and justifying the quality of edges. However, they do not address the structural and functional accuracy of edges in images with a wide range of natural variabilities. In this review, we provide an overview of all the most relevant performance metrics that can be used to benchmark the quality performance of edges in images. We identify four major groups of metrics and also provide a critical insight into the evaluation protocol and governing equations.

Performance of compressed sensing for fluorine‐19 magnetic resonance imaging at low signal‐to‐noise ratio conditions

“…Fluorine ( 19 F) MRI has been commonly used in cell tracking and molecular imaging. [1][2][3][4] Its widespread applicability, along with the challenge that 19 F is only sparsely available in vivo, has motivated the development of strategies that enhance signal-to-noise ratio (SNR) efficiency, for example, more sensitive hardware [5][6][7] and more time-efficient pulse sequences. [8][9][10] Digital signal processing, particularly compressed sensing (CS), [11][12][13] is a new avenue to boost detection performance.…”

“…The first application of CS to 19 F MR methods was in chemical shift imaging. [14][15][16] When applied to 19 F MRI, the potential of CS to improve acquisition time efficiency was demonstrated. 17,18 Under optimal conditions, CS improved SNR efficiency by a factor of ~8.5.…”

“…17 In a phantom experiment, reinvesting the time saved by undersampling into increased averaging lowered the detection threshold, when compared to fully sampled data at equal acquisition time. 18 Application of CS to in vivo 19 F MR was previously shown in situations where the location of 19 F signals in vivo was known: localized wound inflammation and transplantation of labeled pancreatic islets. 17,18 In the latter, 4-fold undersampling lowered the detection threshold, but also introduced false positives.…”

“…It is these conditions which necessitate sensitivity improvements most to boost detection and make high resolutions viable. SNR efficiency does not convey information regarding edge preservation, 19,20 and image defects are generally not captured by known noise distributions. 17 Given that CS is a nonlinear algorithm, [11][12][13] results cannot be extrapolated from experiments studying higher SNRs, and reconstruction methods must be compared at equal acquisition time.…”

“…17 Given that CS is a nonlinear algorithm, [11][12][13] results cannot be extrapolated from experiments studying higher SNRs, and reconstruction methods must be compared at equal acquisition time. 21 This is particularly relevant for quantitative 19 F MR studies and raises a fundamental question under which conditions CS improves detection performance without compromising data fidelity. To draw valid conclusions about performance at low SNRs, a large number of reconstructions, signal distributions, and sampling patterns must be investigated to exclude random effects.…”