Evaluation of Keypoint Detectors and Descriptors in Arthroscopic Images for Feature-Based Matching Applications

Marmol, Andres; Peynot, Thierry; Eriksson, Anders; Jaiprakash, Anjali; Roberts, Jonathan; Crawford, Ross

doi:10.1109/lra.2017.2714150

Cited by 18 publications

(12 citation statements)

References 27 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We express the saliency of arthroscopic images in terms of the number of matched Scale-invariant-feature-transform (SIFT) [10] and Speeded up Robust Features (SURF) [11] key-point features. SIFT and SURF are the state-of-the art feature detection and description algorithms and they have been deployed in arthoscopic environment, with SIFT being the most successful [12]. RANSAC is the state-of-the-art algorithm for outlier rejection commonly used in visual navigation.…”

Section: A Visual Saliencymentioning

confidence: 99%

“…The literature provides solutions for the uncertainty estimation in the internal knee joint measurement [26] and real-time joint motion analysis have been proposed, but they do not address the problem of the autonomous visual navigation. To this end, in the context of visual navigation, stateof-art feature detectors and descriptors for monocular kneearthroscopic images were investigated [12]. Results showed that SIFT features could be best extracted and matched (compared to SURF and others) in knee-arthroscopy, but the study used only sequences containing six unrealistic images not representative of the complexity and length of the procedure.…”

Section: Surgical Visionmentioning

confidence: 99%

“…As such, we approach the problem from a different perspective than the state-of-the art in arthroscopy [12] and consider the reason for our images to be so difficult to extract and subsequently track salient features. We take into consideration the underwater light degradation and then address the problem of feature-poor images by investigating contrast enhancement techniques.…”

Section: Surgical Visionmentioning

confidence: 99%

See 2 more Smart Citations

Saliency Improvement in Feature-Poor Surgical Environments Using Local Laplacian of Specified Histograms

et al. 2020

Self Cite

View full text Add to dashboard Cite

Navigation in endoscopic environments requires an accurate and robust localisation system. A key challenge in such environments is the paucity of visual features that hinders accurate tracking. This paper examines the performance of three image enhancement techniques for tracking under such feature-poor conditions including Contrast Limited Adaptive Histogram Specification (CLAHS), Fast Local Laplacian Filtering (LLAP) and a new combination of the two coined Local Laplacian of Specified Histograms (LLSH). Two cadaveric knee arthroscopic datasets and an underwater seabed inspection dataset are used for the analysis, where results are interpreted by defining visual saliency as the number of correctly matched key-point (SIFT and SURF) features. Experimental results show a significant improvement in contrast quality and feature matching performance when image enhancement techniques are used. Results also demonstrate the LLSHs ability to vastly improve SURF tracking performance indicating more than 87% of successfully matched frames. A comparative analysis provides some important insights useful in the design of vision-based navigation for autonomous agents in feature-poor environments.

show abstract

Section: A Visual Saliencymentioning

confidence: 99%

Section: Surgical Visionmentioning

confidence: 99%

Section: Surgical Visionmentioning

confidence: 99%

See 1 more Smart Citation

Saliency Improvement in Feature-Poor Surgical Environments Using Local Laplacian of Specified Histograms

et al. 2020

Self Cite

View full text Add to dashboard Cite

show abstract

“…4) To measure the instrument gap, a requirement is to accurately track the anatomical points of interest inside the knee joint that forms the instrument gap, which is beyond the scope of this study. Marmol et al evaluated options for tracking feature inside the knee joint [8].…”

Section: B Assumptionsmentioning

confidence: 99%

Robotic Arthroscopy: The Uncertainty in Internal Knee Joint Measurement

et al. 2019

Self Cite

View full text Add to dashboard Cite

We report a study that developed algorithms to measure the dimension and uncertainty range of free space inside the knee joint for the purpose of minimally invasive surgery. During knee arthroscopy, the patient's leg position is continuously adjusted to create the space for surgical instruments inside the joint. Surgeons 'feel' the force they apply to the leg and estimate the joint space from a 2D video. In many cases, they overestimate the instrument gap, resulting in damaging to the knee joint by pushing instruments through a gap that is too small. We used cadaveric experiments to inform the noise induced by the sensors and image processing steps, to derive an error point-cloud in a simulated environment. From the point-cloud, we calculate the instrument gap range inside the knee joint. For a selected surgical instrument gap size, the measurement algorithm is accurate to less than a millimetre. However, measurement errors introduce an uncertainty of 14%. The performance of our algorithms demonstrates the use of a single-lens arthroscope to measure the instrument gap to provide feedback to a surgeon or enable control of a robotic leg manipulator.

show abstract

“…Evaluations of both performance and quality of different key point detection methods were discussed [25][26] . Finally; Merging several key point detection techniques is useful if it is handled well [24,27] .…”

Section: Introductionmentioning

confidence: 99%

Advanced selective key point detection techniques

Amein

El-Tanany²

2019

Eng. Sci. and Milit. Techno.

View full text Add to dashboard Cite

SAR imaging needs precise registration process to get recompensation or setoff between two or more images for a certain area. Generally, operational goal of registration process is a geometrical transformation for the input leading to geometrically agreement for input images. Coregistration step can be used in a lot of fields as change detection, mosaicking, super-resolution images creation, … etc. Registration process can be classified into two main classes: Classical or Automatic methods. For the classical approach, the anchor points are picked out manually and then the transformation modeling is performed, that leads to more time consuming and low output accuracy. Automatic detection of these anchor points aims to recover the functionality of non-automated processes. Registration process should recover many problems such as illumination changes, variations of intensity, Different sensors, noise, … etc. Its applications are mainly dependent on errors, such as multi temporal; multi view; or multi modal, occurred during capturing process. Feature detection, as a step of image registration process, obtains a group of anchors, that are stable and informativeness, or patches under varying conditions. Also, it is critical for the detector to have robustness against viewpoint changes, brightness changes, and other deformations. Detailed discussion and investigation are paid, in this manuscript, for the common corner detectors helping to be familiar with the various applied techniques for feature detection. Our goal is to examine a combination hypothesis of these methods into an advanced selective one.

show abstract

Evaluation of Keypoint Detectors and Descriptors in Arthroscopic Images for Feature-Based Matching Applications

Cited by 18 publications

References 27 publications

Saliency Improvement in Feature-Poor Surgical Environments Using Local Laplacian of Specified Histograms

Saliency Improvement in Feature-Poor Surgical Environments Using Local Laplacian of Specified Histograms

Robotic Arthroscopy: The Uncertainty in Internal Knee Joint Measurement

Advanced selective key point detection techniques

Contact Info

Product

Resources

About