Multi-spectral and Thermography Imaging Techniques for the Investigation of a 15th Century Wall Painting

: When planning the restoration of an artwork, the good practice involves the evaluation of


Introduction
In developing restoration, conservation and preservation plans for Cultural Heritage (CH) items, diagnostics is nowadays a key ingredient.The knowledge of both the original and restoration materials, of the construction techniques, and of the stratigraphy of the artwork is requested by conservators to define any potential interventions.Among numerous diagnostics tools, imaging techniques have become widely applied in CH as they provide information on the materials and the structure of the artworks in a non-invasive and non-destructive way.Such information is useful for conservators to choose the most appropriate intervention decision during a restoration work [1][2][3][4][5][6][7][8][9][10][11][12][13][14][15][16].
Imaging techniques can provide a complete knowledge of the surfaces in a fast and reliable way without the need for sampling micro-chips of materials for laboratory analysis, a fact that is in general highly desirable or mandatory in some cases [13][14][15][16].
When dealing with paintings, a thorough diagnostic approach relies on (i) the integration of data collected from different imaging techniques, (ii) the exploitation of their complementarities in terms of spectral range and/or physical principles, and (iii) the potential combination of the relative outputs via digital imaging processing tools.It is important to note that data integration and fusion can strongly support traditional analytical methods, yielding a comprehensive diagnosis of the painting state of conservation and reducing any ambiguities due to the use of a single diagnostic method [2,[9][10][11]14,[16][17].
Hyperspectral (HS), multispectral (MS), and any other imaging techniques processing spectral data are the preferred ones for studying the outer layers of painting such as the varnish, the pictorial stratum, the dirt -if present -and the underdrawings.Thus, starting from ultraviolet (UV), passing by visible (VIS), and reaching infrared (IR), the outer layers of a painting can be inspected.In particular, UV reflectance and fluorescence are employed for evaluating the condition of the varnish layer, for identifying possible retouching and over-paintings; fluorescence can also aid in the characterization of the materials (varnish, pigments and binders).The analysis of the reflected light in the VIS range, especially using raking light, highlights the artist's technique (brushstrokes, impastos), the presence of flaking paint or cracks, and it can also show areas requiring special care for being faithfully preserved.Finally, images in the near IR range are used for the analysis of the inner layers -near IR light can penetrate the pictorial layer and make the underdrawing and possible pentimenti being visible due to the high reflectance of carbon-based materials.This is possible since most of the historical pigments, binders, and varnish are partially transparent in the near IR (around 1000 nm of wavelength).However, the transparency of the outer layers to the near IR lights depends on many factors such as the layer thickness and chemical composition, a fact that makes the imaging of the underdrawing layer challenging.Another powerful exploitation of the IR radiation for painting analysis is represented by the HS imaging techniques that use the retrieved pixelwise spectral information to classify the pigments and to map them over the inspected painting surface.Hyperspectral cameras split the range of sensitivity of the sensors in hundreds of bins for an accurate reconstruction of the reflected IR spectrum.In painting inspections, near (900-1700 nm) or short-wave (1000-2500 nm) IR hyperspectral camera are mainly used -many of the overtones and the combination of vibrational bands characteristics of the inorganic pigments used in historical paintings fall in these ranges.It should be noted that the application of HS imaging in the middle (3000-5000 nm) IR has also been recently reported to classify organic compounds [18].
The main drawback of HS imaging is a practical one: hyperspectral cameras are still very expensive and need for a scanning stage to output images of the surface -the spectral scanning is usually realized by combining a standard IR camera with a monochromator or with a diffraction grating.These facts hamper the use of HS imaging for in situ artworks inspection.Further, for what mentioned above, HS imaging should be anyway combined with UV and VIS imaging methods.As a consequence, the possible engineering of a single technique capable of providing info within the mentioned electromagnetic spectra is highly desirable and would be of a great value from a practical point of view.In this framework, HMI processes information from the near UV to the near IR.The here-employed system relies on a 36 megapixels reflex Nikon camera modified to work in the 300-1000 nm extended range combined with proper choice of flash lamps and optical filters.The images obtained are then represented into a 7D hyper-colorimetric space.Such coordinate system extends the concept of the common 3D colour space used in colorimetry, i.e. the CIE XYZ standard, or the more common RGB, by defining seven hyper-colorimetric coordinates in which a coordinate is related to the near UV range and three ones to the near IR range.By combining artificial intelligence techniques with a proper calibration procedure, it is possible to fully exploit this augmented colour space to perform all the tasks discussed above [19][20][21].
Regarding the evaluation of the whole painting stratigraphy, i.e. from the varnish to the support (panel, canvas, wall, etc..), the number of possible techniques is even larger: X-ray imaging is largely employed, 2D and 3D XRF imaging are often used for identifying materials' chemical composition.In addition, medical computer tomography techniques, terahertz time resolved (TTR) imaging, optical coherence tomography (OCT) and nuclear magnetic resonance have recently enlarged the range of methods for such purpose [22][23][24].
In this framework, InfraRed Thermography (IRT) represents a robust technique for the inspection of CH items [25][26].Also in this case, a trade-off between information amount, cost and measurement complexity (measurement time, mechanical equipment, computational costs, etc.) should be found.IRT is relatively easy to be performed in situ, providing imaging without the need of imaging reconstruction algorithms and it can be operated on large areas moderately easy.Moreover, IRT does not require a continuous translation stage as for the point-wise imaging equipment such as TTR, XRF, OCT, NMR, etc., and it does not involve any hazardous source.The costs are not as low as the typical equipment for IR reflectometry, but are among the cheapest ones within the above-mentioned techniques.
On the other hand, IRT does not reach the depth resolution of other techniques such as TTR and OCT, but it can inspect the support up to depth of several millimetres -a few centimetres in the case of wall paintings provided that a suitable excitation scheme is used.In addition, recent developments in IRT technique have increased its effectiveness for artworks inspection and reduced the risk of any alteration of the artworks due to the thermal stimulus [27][28].In particular, the use of the so-called Pulse-Compression Thermography (PuCT) procedure combined with a pseudo-noise modulated heating stimuli demonstrated to be effective even with a very small increment in the sample surface's temperature (∼1°C) [29][30].
For the mentioned reasons, the combined use of PuCT and HMI is very promising for in situ inspections of paintings: the tailored hardware and the advanced processing and postprocessing algorithms provide complimentary information to art historians and restorers, and their contextual use allows for investigating the paintings' stratigraphy from the varnish to the support.The output of the two methods can be easily fused by image processing tools and post-processing algorithms.
The first example of the combined application of the two techniques was recently reported in the literature -HMI and PuCT were used for inspecting two historical panel paintings of the renaissance period, realised by Andrea Mantegna and from Michelangelo's workshop [31].
In the present work, HMI and PuCT are applied together for the first time on a detached wall painting under restoration in the Laboratories of the five-year course in Conservation and Restoration of Cultural Heritage (LMR/02) of University of Tuscia, Italy [32].
The wall painting, specifically a lunette, depicts a Madonna and Child enthroned in between the angels, and the Saints Jerome and Francis (Fig. 1).The painting has been dated back to 1490 and it is attributed to the artist Antonio del Massaro, known as Pastura (1450-1519).Originally, it was in the convent of Santa Maria del Paradiso in Viterbo, central Italy [33][34].
After the detachment occurred in 1912, the painting was transferred in the Civic Museum of Viterbo during the opening day of the new seat in Santa Maria della Verità church.The detachment of the lunette was probably due to its bad state of conservation caused by the exposure to weathering and acts of vandalism [35].In fact, large grouts are visible in the photographs of the beginning of 20 th century confirming its bad state of conservation and the need for interventions [36].
The ancient restoration interventions were made with both unsuitable materials and techniques such as the mimetic reintegration, hampering in turn the correct readability of the artwork and of its original painting.In accordance with the Superintendence and the responsible of the Civic Museum, it has been thus decided to perform a restoration aimed at recovering the original appearance of the painting at the best.Consequently, in 2016 the lunette has been brought in the restoration laboratories of University of Tuscia for starting the intervention.
The restoration work has been initially supported by traditional ultraviolet fluorescence (UVF) photography aimed at mapping the superimposed materials, the grouts, the lacunae and in general the non-original areas being very well visible under ultraviolet fluorescence photography.The image under UV radiation showed diffuse blue fluorescence probably due to the presence of glue used for the detachment of the lunette [36][37].Grouts and cracks have been also well-highlighted by the UV image [32].Moreover, in order to characterise the blue fluorescent material visible under UV radiation, Fourier transform infrared analysis was performed.This revealed that the yellowed surface material was made of protein glue.The main signatures of glue have been observed in the infrared spectrum at cm -1 : 3289, 3064, 2930, 1660, 1542, 1373, 1313, 1246, 1033 [38][39].Other bands in the spectrum have been associated to calcium carbonate (1424 and 876 cm -1 ) and gypsum (1152 cm -1 ) [32].This preliminary analysis has been fundamental both to address the first cleaning choices of restorers and to start the intervention.
After this, further investigations were performed by combining HMI and PuCT to obtain information about the painting materials and the stratigraphic sequence, the last one being a highly-inhomogeneous mix of both original layers and new grounds added during the detachment of the lunette from the church of Santa Maria del Paradiso.The achieved results are illustrated in the following sections, demonstrating the high potential of this integrated approach to become a standard methodology in the preliminary study of the artworks to be restored.

Hypercolorimetric Multispectral Imaging (HMI)
HMI was performed through a Nikon D810FR 36 Megapixel camera, modified to obtain full range spectral reflectance measurements.Nikon SB910 xenon flashes without their front plastic lenses were used for lighting the painting, thus allowing the UV wavelength to be emitted as well.
The UVF was then obtained by filtering the flashes light with a UV band-pass filter with a spectral cut at 380 nm, and UV-IR cut filter (400-700 nm) in front of the camera.The HMI image processing system consists of two main software tools, i.e.SpectraPick ® for the image calibration and PickViewer ® for the image analysis [13,[19][20][21]31].The calibration of the images was achieved by placing radiometric references just below the painting.In particular, various white patches and a sample with 36 patches of colour-checkers built using colour samples from the NCS -Natural Colour System ®© catalog were placed just below the painting, see Fig. 1.The spectral reflectance of the references was measured in the range 220-1050 nm at Profilocolore laboratory, with 0.7 nm accuracy (Instrument System Spectroradiometer CAS 140 CT and dark room).The calibration procedure outputs a single AdobeRGB TIFF 16 bit colour image and seven monochromatic images in 16 bit TIFF format, containing the spectral reflectance values at 350, 450, 550, 650, 750, 850 and 950 nm.The last three values were used to define the IR1, IR2 and IR3 images respectively.The precision achieved in the reflectance measurement across the whole 36 megapixels image is higher than 95% and the colour error less than CIE2000 ∆E=2 for the colour image [13].The whole calibration and alignment process require a few minutes and it can be performed in situ for an immediate results analysis.
After the image acquisition and calibration, multispectral images were processed through the HMI software PickViewer ® , developed by Profilocolore.The PickViewer ® software provides powerful image processing tools able to reveal relevant information.Further, it allows gathering information on hidden data from the images acquired with SpectraPick system, containing spectral reflectance and colour coordinates for each of the 36 megapixels of the captured scene.Data are calibrated and they are absolute in values, thus only depending on the spectral characteristics of the surface.Several kinds of analyses are possible with PickViewer ® , such as: -adding and integrating any other imaging data (fluorescence, X-ray, thermal etc.); -multichannel images viewer; any pixel colorimetry and spectral reflectance read-out; -mapping by colour, spectra, arbitrary channels; Each relevant result can be saved as image in TIFF, png or jpeg format.Results saved in TIFF format can be reloaded as further derived channels and used combined with all the others [19].

Pulse-Compression Thermography (PuCT)
The experimental setup used for the PuCT is the same detailed in previous published papers, wherein also an extended theoretical base of the technique can be found [32][33][34].In particular, the signal generation/acquisition was managed by Labview TM software.Thermograms were collected through a Xenics Onca-MWIR (3.6-4.9 μm)-InSb, having a resolution of 320×240 pixels, a frame rate of 40 Hz, and connected to a NI-1433 Camera Link Frame Grabber.The distance between the painting and the camera was about 70 cm.Eight light emitting diode (LED) chips have been used as the heat sources, each one with a maximum nominal electrical power of 50 W.Note that the LEDs were not used at their maximum power -an overall electrical peak power of about 250 W was reached to get the reported results.The pseudo-noise code driving the LED chips consisted in a repetition of two Legendre sequences, each one of 47 bits.The duration of each bit was 1 s, thus resulting in an overall excitation of 94 s.During this time, the LEDs were switched on for about half of the overall excitation time, so that an overall electrical energy of ∼12 kJ was delivered.Note that the camera and the heating source were on the same side with respect to the inspected painting, thus the PuCT was performed in reflection mode.The coded excitation voltage signal driving the LEDs was provided by a TDK Lambda GEN 750W power supply.The frame grabber and the power supply were synchronously driven by the signals provided by a National Instrument PCI-6711 Arbitrary Waveform Generator (AWG) board.Both the AWG board and the frame grabber were connected to a central PC/DSP Unit.
A thorough description of the PuC algorithm lies beyond the scope of this work, therefore the reader is referred to [28,[30][31] for a deeper understanding of the whole procedure.For the sake of clarity, some fundamentals aspects are worth to be here recalled: 1) For the here-employed setup, the reconstructed time trend of each pixel after the application of the PuC algorithm corresponds to that achievable by exciting the sample with an equivalent rectangular pulse of 1 s duration.The advantage of using PuC instead of a pulsed excitation is that this fictitious single pulse carries the energy of the entire pseudo-noise Legendre sequence, so that the response after PuC, and hence the thermograms, are characterized by a good signal-to-noise ratio (SNR) even by using low peak power values.It must be stressed that such aspect is crucial in the inspection of irreplaceable items as the artworks: while a short high-power excitation pulse could damage the painting by inducing thermochromism or thermal/mechanical stress, the use of a pseudo-noise excitation spread instead the energy within a longer time, making the temperature rise of the sample being smooth and lower in magnitude.After PuC, a fictitious larger and faster temperature increment is instead achieved exploiting the mathematical properties of the coded excitation.
2) It has been shown in [31] that the use of the Hilbert transform and of the derived time-phase feature from PuCT signals can help to highlight some characteristics of the support, such as the wood grain, grouts, etc.. Here, the same processing is applied to the PuCT data to enhance the sensitivity in the inspection of sub-surface layers.Thus, three different time-domain features will be here shown.In particular: "Emissivity" -is the amplitude of the time signal of each pixel retrieved after PuC and it is an indirect measure of the pixel thermal emissivity, which can be related to the pixel surface temperature by a proper calibration; "Hilbert" -is the output of the Hilbert transform applied pixelwise to the "Emissivity" trends; "Time-phase" -is the function that for each time instant  outputs the argument of the complex number () = () +  * () in time domain, with i being the imaginary unit.
3) As for HMI, PCA has been applied to the time series of thermograms produced by imaging both the Emissivity and the Hilbert features.As can be seen in the next Section, PCA allows for evidencing some specific characteristics of the time-series while reducing the number of images to be analysed.This can be very useful for a first analysis of the sample to detect anomalies, while for a quantitative analysis the processing of the image time-series is necessary.4) An alternative way for presenting the thermography results has been here explored, which is based on using the three abovementioned features, i.e. emissivity, Hilbert and time-phase, to generate colour images.In particular, RGB, and YCbCr colour spaces have been explored by relating each colour coordinate to the each of the three features.
As a final remark, it is important to highlight that the PuCT images showed below were obtained after merging thirteen thermal acquisitions.Multiple acquisitions were performed to obtain a good trade-off between the tested painting surface and the need for assuring enough spatial resolution of the captured thermograms.As an example of the procedure, Figure 9a depicts the image of the thermal emissivity estimated after 1 second from the beginning of the reconstructed thermal pulse.Some discontinuities between the images corresponding to the various acquisitions can be seen even after some preliminary image processing.Certainly, the merging procedure can be further improved with a proper calibration, but this will be addressed in a next work.
The current results are anyway enough satisfactory to be exploited for an evaluation of the state of conservation of a painting.

Hypercolorimetric Multispectral Imaging (HMI)
The seven calibrated spectral bands were analysed in PickViewer ® software and some processing tools were applied to the images in order to gain further diagnostic information.
The results of the processing are shown in Fig. 2-8.PCA was applied on IR images to highlight the presence of preparatory drawing and possible pentimenti, especially in the restored areas where different and superimposed materials are found to be highly mixed.PCA has been widely used for art conservation applications, either as a stand-alone technique or to reduce the dimensionality of datasets using linear algebra techniques prior to another classification algorithm whereby multiple variables are analysed to evaluate the contribution of each variable to an observed result [40][41].
Clustering techniques can classify data based on a comparison of the spectral character of each point in the image with the spectral character of every other point; this permits patterns or clusters to emerge from the clustered data, thus pointing out areas of compositional similarity and returning a spatial distribution of components [42].In the detail of the foot of Saint Jerome in the IR2 image, a pentimento can be observed (Fig. 3) on the left side, where a different original outline was traced for drawing the foot.Figure 3 shows also the graphical user interface (GUI) of PickViewer ® software: on the left an image or a selected area of the image is shown, while the processing results, according to the selected function, are visible on the right part.Extensive micro-cracking has been highlighted in correspondence of the Virgin's face by applying PCA to the images in the IR region (IR1, IR2 and IR3, see Fig. 4).
Another interesting processing tool of Pickviewer ® is Spectral Similarities that allows for comparing the spectral reflectance of two different points on the examined surface.It was applied, for example, to the red colour in the Virgin' garment and in that of the angel at the right side of the Virgin (Fig. 5).The values of colour coordinates and of reflectance show the high similarity of the two point colour, suggesting a possible equal composition in terms of pigments.
In order to make hypothesis about pigment composition, especially for the blue and green ones constituting the background and the Virgin mantle, infrared false colour (IRFC) analysis has been obtained by combining calibrated R, G, B and IR channels (Fig. 6) [43][44][45].The possibility of obtaining the IRFC image in a very short time by simply combining the channels, is a great potentiality of PickViewer ® software.For the blue pigment of the Virgin's garment, the use of azurite was hypothesized on the base of its characteristic deep blue colour in IRFC.The sky and parts of the vest of the Virgin have a particular green colour in the visible; in IRFC, they assume a light blue hue making probable its attribution to malachite.Considering the bad state of conservation of the artwork after a long exposition to weathering, a preferential degradation caused by humidity with a transformation of azurite into malachite or other green copper-based compounds seems very likely, rather than the application of a green pigment in areas meant to be blue-coloured [46].Portion of unaltered azurite are in fact clearly visible in the background sky behind the head of the Virgin and in the lower edge of the sky as well (Fig. 7).In order to investigate this possibility in a non-invasive way, a map of spectral similarity was built on the base of the 7-bands curve of spectral reflectance measured in a 9×9 pixels area in the blue part of the Virgin's dress, as shown Fig. 8.The output showed on the right of the visible image in the software's GUI, highlights a correspondence between the blue pigment of the vest of the Virgin and the one used to paint the sky, supporting the hypothesis of the use of azurite in all the blue areas.
The painting layer of azurite used for the sky is applied over a coloured ground layer probably based on red earth (hematite) or charcoal black (morellone), usually used for the application of azurite in wall paintings, as described in scientific literature [47].

Pulse-Compression Thermography (PuCT)
The reported results were obtained by applying pixelwise the Pulse-Compression (PuC) algorithm over the acquired thermograms, which has been described in detail in Sect.2.2.
The IR2 band (850 nm, Fig. 2) gives some information on the preparatory drawing of the scene, as it is visible in the garments of the saints, of the angels, of the Virgin and of the lion near to Saint Jerome.Concerning the infrared PuCT analysis, the results highlight many interesting details of the surface and of the stratigraphy.Detachments, gold coating and different restoration works appear as brighter areas in emissivity images.Note that the brighter the area, the lower the emissivity.This can be seen in Fig. 9 reporting the emissivity image corresponding to  = 1 s, at which the fictitious excitation pulse stops, and the temperature decay starts.Typically, this is the time instant that better shows the details of the pictorial layer, since during the excitation the light is absorbed differently by the various colours, leading to a variation of the surface temperature over the sample.
The presence of gold traces in the haloes of the saints and angels, suggests that the haloes were originally gilded probably by using a missione, an adhesive oil/resin primer, used to make the adhesion of gold on the wall paintings possible [48], visible in the IRFC image with a yellow-brown colour.
Among the various areas of interest, the large previous restoration work between the Virgin and Saint Jerome appears as a well-delimited cold area, while Saint Jerome's right arm is hotter than the rest of the figure.It can be noticed that the same arm exhibits an anomalous brilliance in the IR2 and in false colour images (Figures 2 and 6), probably due to the use of different materials and painting techniques.
Then, different features become more visible as time elapses, as there is a direct link between the thermal responses of layers at deeper depths within the tested materials and the flow of time.9, four lighter-colder spots are clearly visible while in Fig. 10 they are no more present but a darker-hotter spot appeared in the middle of the four previous spots.The evolution of this area in time is illustrated with more details in Fig. 11, where a zoom of this part is reported as time elapses together with the visible image serving as a reference.
To better identify the support of the paintings, the images obtained by Hilbert and time-phase features can be used.Indeed, while the emissivity feature is influenced for a quite long time by the temperature rise of the pictorial layer, in the other two features, the details of the pictorial layer, i.e. the painting subject, disappears rapidly and the subsurface layers can be visualized.Other frames corresponding to different time instants could be selected to better highlight different details, but the main aim is here to demonstrate the capability of the PuCT technique to provide information from the inner structure of the sample to be complemented by HMI information.
PCA can be applied separately to emissivity, Hilbert and time-phase images, or to all the images together.It has been found that the application of PCA to the emissivity time series only allows most of the details of interest to be visualized in a few images, as it can be seen in Figures 13 to 15 showing the first three PCA components.In particular, the PCA3 image in Fig. 15 depicts the canvas under the painting, even if these were not so appreciable in the emissivity images.
As a further investigation, the data reported in Figs.13-15 have been transformed from the grey scale to the YCbCr colour space [49].This is because the human eye have a different perception between a grey scale image and its coloured replica, thus this can help in highlighting areas having different thermal contrasts.Although the perception is inherently highly subjective by nature and further investigation are needed to confirm the validity of the proposed approach -including the possibility to submit a survey to several experts in the field -in our opinion, Figs.The combined use of the two techniques, for the first time on a wall painting, was possible thanks to the restoration work, offering the chance for investigating the artwork to study the materials, the construction techniques, the state of preservation and the stratigraphic pattern as consequence of the detachment operations performed at the beginning of the 20 th century.HMI supplied relevant information about surface and sub-surface layers in terms of possible pigment composition and distribution, and preparatory drawing/pentimenti, respectively.
PuCT output signatures of detachments, grouting, gilding from the surface to the deep layers giving relevant information about the possible presence of discontinuity or deep grouts.A color space transformation from gray scale to YCbCr was found to be useful for highlighting areas of interest.
Further processing of the acquired images will be possible also with the support of conservators that could address the choice of the most useful deepening to supply a valid aid to the intervention.

Declarations
The  The HMI infrared calibrated image at 850 nm (IR2) showing the preparatory drawing used for the construction of the painting.PuCT Hilbert image obtained at =2.5 s.

Figure 1 .
Figure 1.The lunette with Madonna and the Child enthroned between the angels and the Saints

-
principal components analysis (PCA); -contrast enhancement through digital imaging processing algorithms; -neural network based clustering; -colour and spectral signature database; -two ways mapping by database entry; -any channel to RGB false colours visualization; -channels math, indexes and normalised contrast; -calibration and colour-checker test.

Figure 2 .
Figure 2. The HMI infrared calibrated image at 850 nm (IR2) showing the preparatory drawing used for

Figure 3 .
Figure 3.A detail of the HMI infrared calibrated image at 850 nm (IR2) showing the pentimento in the

Figure 5 .
Figure 5. Spectral Similarities tool applied to a point on the arm of the Virgin garment and on the Angel's

Figure 7 .
Figure 7. Detail of IRFC image showing the blue portion of the probable unaltered azurite in the

Figure 10 .
Figure 10.PuCT emissivity image obtained at  = 7 s.As an example, Fig. 10 depicts the emissivity image at  = 7 s.Signatures of possible deeper

Figure 11 .
Figure 11.PuCT emissivity images at different times representing Virgin's face and cloak closure.Darker

Figure 12
Figure12shows the image of the Hilbert feature at  = 2.5 s, where the canvas texture -including

Figure 13 .
Figure 13.First PCA image retrieved by PuCT emissivity time series image.

Figure 14 .
Figure 14.Second PCA image retrieved by PuCT emissivity time series image.

Figure 15 .
Figure 15.Third PCA image retrieved by PuCT emissivity time series image.

Figure 16 .
Figure 16.False-colour PuCT -YCbCr colour space at t =1 s PCA image retrieved by PuCT emissivity time

Figure 17 .
Figure 17.False-colour PuCT -YCbCr colour space at t =3 s PCA image retrieved by PuCT emissivity time

Figure 18 .
Figure 18.False-colour PuCT -YCbCr colour space at t =4 s PCA image retrieved by PuCT emissivity time (HMI) and Pulse-Compression Thermography (PuCT) have been tested on a detached wall painting representing the Madonna and the Child enthroned between the angels and the Saints Jerome and Francis (AD 1490), attributed to the painter Antonio del Massaro known as Pastura (1450-1519).
lunette with Madonna and the Child enthroned between the angels and the Saints Jerome and Francis (AD 1490), attributed to the painter Antonio del Massaro known as Pastura (1450-1519).The calibrated visible image with the colour checker and the white patches used for HMI acquisition is shown in the gure.

Figure 3 A
Figure 3

Figure 7 Detail
Figure 7

First
PCA image retrieved by PuCT emissivity time series image.

Figure 14 Second
Figure 14

Figure 15 Third
Figure 15

Figure 16 False
Figure 16

Figure 17 False
Figure 17

Figure 18 False
Figure 18