Spatial-Spectral Representation for X-Ray Fluorescence Image Super-Resolution

Dai, Qiqin; Pouyet, Émeline; Cossairt, Oliver; Walton, Marc; Katsaggelos, Aggelos K.

doi:10.1109/tci.2017.2703987

Cited by 11 publications

(15 citation statements)

References 26 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…When the difference between two adjacent RGB pixels is small, a strong spatial smoothness constraint is placed on their corresponding XRF pixels, and vice versa. This adaptive TV regularizer is one of the main differences between this fusion-based XRF image inpainting algorithm and our previous fusionbased XRF image SR algorithm [36]. We found out that such TV regularizer on the visible component is essential for the inpainting problem, otherwise the inpainting results are not satisfactory.…”

Section: B Proposed Solutionmentioning

confidence: 86%

“…In this section, we propose the XRF image inpainting algorithm by fusing it with a conventional RGB image, detailing the "Proposed Inpainting Algorithm" block in Figure 3. The proposed fusion style inpainting approach has similarities with our previous fusion style SR approach [36]. We first formulate the XRF image inpainting problem, then demonstrate our proposed solution to this inpainting problem.…”

Section: Spatial-spectral Representation For X-ray Fluorescence Imentioning

confidence: 98%

“…In addition, the correlation among spectral channels needs to be preserved for the inpainted pixels. In our previous work on spatial-spectral representation for XRF image superresolution [36], the spatial resolution of the visible component XRF signal is increased by fusing an HR conventional RGB image while the spatial resolution of the non-visible part is increased by using a standard total variation regularizer [47], [48]. Here we propose an XRF image inpainting algorithm by fusing an HR conventional RGB image, which can be regarded as an extension of our previous XRF SR approach.…”

Section: Xrf Imaging Using Adaptive Samplingmentioning

confidence: 99%

“…The third term in Equation 14 is a TV regularizer of the non-visible component. Its detailed definition can be found in our previous work [36] (see Equation (10)).…”

Section: B Proposed Solutionmentioning

confidence: 99%

“…de Heem's "Bloemen en insecten" (ca 1645), in the collection of Koninklijk Museum voor Schone Kunsten (KMKSA) Antwerp, are presented here. We utilize the super-resolved XRF image in our previous work [36] as the ground truth. The ground truth XRF image has dimensions 680×580×2048.…”

Section: B Adaptive Sampling Mask For X-ray Fluorescence Image Inpaimentioning

confidence: 99%

See 4 more Smart Citations

Adaptive Image Sampling Using Deep Learning and Its Application on X-Ray Fluorescence Image Reconstruction

Dai

Chopp

Pouyet

et al. 2020

IEEE Trans. Multimedia

Self Cite

View full text Add to dashboard Cite

This paper presents an adaptive image sampling algorithm based on Deep Learning (DL). It consists of an adaptive sampling mask generation network which is jointly trained with an image inpainting network. The sampling rate is controlled by the mask generation network, and a binarization strategy is investigated to make the sampling mask binary. In addition to the image sampling and reconstruction process, we show how it can be extended and used to speed up raster scanning such as the X-Ray fluorescence (XRF) image scanning process. Recently XRF laboratory-based systems have evolved into lightweight and portable instruments thanks to technological advancements in both X-Ray generation and detection. However, the scanning time of an XRF image is usually long due to the long exposure requirements (e.g., 100µs − 1ms per point). We propose an XRF image inpainting approach to address the long scanning times, thus speeding up the scanning process, while being able to reconstruct a high quality XRF image. The proposed adaptive image sampling algorithm is applied to the RGB image of the scanning target to generate the sampling mask. The XRF scanner is then driven according to the sampling mask to scan a subset of the total image pixels. Finally, we inpaint the scanned XRF image by fusing the RGB image to reconstruct the full scan XRF image. The experiments show that the proposed adaptive sampling algorithm is able to effectively sample the image and achieve a better reconstruction accuracy than that of existing methods.

show abstract

Section: B Proposed Solutionmentioning

confidence: 86%

Section: Spatial-spectral Representation For X-ray Fluorescence Imentioning

confidence: 98%

Section: Xrf Imaging Using Adaptive Samplingmentioning

confidence: 99%

“…The third term in Equation 14 is a TV regularizer of the non-visible component. Its detailed definition can be found in our previous work [36] (see Equation (10)).…”

Section: B Proposed Solutionmentioning

confidence: 99%

Section: B Adaptive Sampling Mask For X-ray Fluorescence Image Inpaimentioning

confidence: 99%

See 3 more Smart Citations

Adaptive Image Sampling Using Deep Learning and Its Application on X-Ray Fluorescence Image Reconstruction

Dai

Chopp

Pouyet

et al. 2020

IEEE Trans. Multimedia

Self Cite

View full text Add to dashboard Cite

show abstract

OXIRIS project: Development of a new XRF device for the simultaneous detection and treatment of cancer

et al. 2021

View full text Add to dashboard Cite

A novel and suitable energy‐dispersive X‐ray fluorescent (EDXRF) device, currently at the prototype stage, the Project OXIRIS (Orthovoltage X‐Ray–Induced Radiation System), is presented for the simultaneous detection and treatment of cancer diseases. Monte Carlo simulation and experimental results of EDXRF signals from a small deep artificial tumor consisting of a solution of gold nanoparticles bio‐targeted and immersed in a tissue‐bioequivalent matrix are presented and compared. Briefly, the device consists of a dynamic orthovoltage X‐ray fluence concentration coupled to confocal with an EDXRF system along with a sample holder for 3D scanning; all integrated and controlled by a dedicated software capable of controlling the whole operation functionalities: the X‐ray source, the rotating arm, the sample holder, and the detection system. The software also includes dedicated subroutines for X‐ray fluorescent spectra processing to correlate K‐lines signal at each acquisition position with the corresponding high atomic number elements' concentration to produce a 3D distribution according to the user‐defined grid. The confocal configuration ensures that the detected signal comes exclusively from the excited volume as defined by the bulk of the focal spot. Hence, the 3D image is achieved by scanning the sample holder through the movement of the sample‐carrier stretcher moved by step motors in the 3 coordinated axes. The feasibility of the proposed methodology and the design of the prototype have been successfully demonstrated experimentally, targeting gold nanoparticles in water‐equivalent phantoms.

show abstract

Hyperspectral Video Super-Resolution Using Beta Process and Bayesian Dictionary Learning

Ghassab

Bouguila

2021

Advances in Visual Computing

View full text Add to dashboard Cite

Spatial-Spectral Representation for X-Ray Fluorescence Image Super-Resolution

Cited by 11 publications

References 26 publications

Adaptive Image Sampling Using Deep Learning and Its Application on X-Ray Fluorescence Image Reconstruction

Adaptive Image Sampling Using Deep Learning and Its Application on X-Ray Fluorescence Image Reconstruction

OXIRIS project: Development of a new XRF device for the simultaneous detection and treatment of cancer

Hyperspectral Video Super-Resolution Using Beta Process and Bayesian Dictionary Learning

Contact Info

Product

Resources

About