An approach to the detection and recovery of lost grisaille painting.


The grisaille painting ,composed of metal oxides mixed with finely ground colourless glass made into paste and with diverse liquids, was spread with brushes on the pieces of glass and the pigments afterwards were fired onto the glass.

Unfortunately, during the centuries the falling off and loss of grisaille painting has constituted one of the most common problems, due sometimes to imperfections in the firing procedures or flaws in the pigment composition, facts even worsened by particular microclimate conditions ‘in situ’.

Furthermore, re-painting , dirt or incautious cleaning operations contributed to the increasing of the grisaille falling off. Therefore, in front of some stained glasses that partially lost their painting the questions that often come up are how to stop the deterioration in act and if there is a way to regain what it is lost. Whereas there is a solution for the former, as it is possible to slow down the painting falling off thanks to particular protective measures, for the latter, the matter seems to be unsolvable considering that it is hard to recover the original painting.

Yet, if regaining the grisaille painting is obviously impossible, a deeper research and study on the few remaining documents as it is being made in the research in act that is presented in the following pages.

First attempts made during the restoration of the Da via Bargellini ‘Madonna and Child’

The Madonna and Child at Da Via-Bargellini Museum in Bologna was restored in 1990 by the Studio Fenice Laboratory. The loss of the grisaille painting is evident on its internal side, but it is free of posterior cold paintings. Except for a light iridescent patina, the pieces of glass on the external side were well preserved. They had been treated with silver stain (on hair and ornament details), with neither grisaille nor rosa-incarnato to intensify the shading.

Furthermore, in the course of the restoration (1) it was noticed that the loss of paint had not been caused by original defects, as the characteristic printing (or ‘ghost image’) left on the glass (2) when the grisaille was fired in the furnace was easily detected thanks to a grazing illumination.

Some circular surface abrasions were also observed due to repeated (and maybe customary) ‘in situ’ cleaning operations dating back to the previous location of the panel (inserted within a window in a staircase, a place particularly sensitive to vibrations).

In spite of the precarious conditions of the stained glass when it arrived at the restoration laboratory, and the lack of any information about either its origin or its author, it was visibly a piece of valuable make, where the printing under the lost grisaille, for the fineness of the execution, revealed the hand of an outstanding artist who had perfectly mastered the technique. Therefore, it was a particularly interesting sample and suitable for a research aimed at the definition of non-destructive methods in order to detect and reconstruct the lost grisaille painting without any manual intervention.

In fact, other approaches were made in the past on ancient glasses but with unsatisfactory results for the lack of non-destructive, reliable and practical methods. An attempt was carried out by Marchini in 1976. He tried to spread graphite on the uneven surface of the glass. The result was interesting but the operation required the use of ‘enamel’ to fix the powder of graphite, so the method, after the experiment made on the stained glass of S. Maria alle carceri in Prato and on the windows of S. Maria Novella sacristy, was no more used afterwards.(3)

The Madonna and Child – a temporary loan kindly allowed by the institution owning the panel (Opera Pia Da Via – Bargellini) - constituted a particularly interesting sample from where to start a new approach. First attempts were carried out using a TV camera connected to a video monitor and consisting of several shots taken with different light sources. It was observed that more or less significant results were obtained by varying the angle of light, of shot between TV camera and subject, and the general conditions of the experiment.

In that way, both ‘negative’ and ‘positive’ images of the printing left by the grisaille were obtained: in particular, the ‘positive’ images immediately offered the surprising feeling of seeing the lost painting reappear on the monitor without any special adjustment.

An approach to the detection: the image scanning

The encouraging results of the first investigations on the Madonna and Child panel were submitted to Professor C. Pirina, President of C.V.M.A. – Italy, who in 1994 promoted the continuation and development of the research, still in progress, at the Istituto di Ricerca sulle Onde Elettromagnetiche ‘Nello Carrara’ of the C.N.R. in Florence

At first it was observed that an adequate method of detection consists in illuminating the glasses in a darkened room with almost collimated beams, placed symmetrically at an angle with the normal to the panel (30-60-degrees) in such a way to have uniform illumination. A clear image of the lost grisaille printing is observed looking frontally at the piece of glass from the side opposite the projector, avoiding to be reached by the light of the beams (transmission mode). The image is in fact due to scattering, i.e. the partial deviation of light from its local incidence direction caused by the diffraction of light on the prints. The bright figure against a dark background, formed by the scattered light intensity, gives the observer the impression of looking at the negative reconstruction of the original painting. The same bright figure can also be seen looking from the side of the projector (projector mode). To obtain the positive pattern of the image the piece of glass should be seen against a diffuse source, but in this case it is hard to reach a good contrast. Positive images are in fact originated by the local subtraction of energy from the incident direction, which appears as a local apparent attenuation of the image of the source. Yet, in the observation of a wide diffuse source, the light emitted from different source points can compensate for this attenuation and fade away the contrast.

The recording of the images can be performed in several ways. A common image scanner connected to a personal computer turned out to be the most advantageous method. In a n image scanner, the relative disposition of the lamp and sensor is such as to avoid the saturation of the sensor with the light reflected from the document supporting glass, which is exactly the right condition to scan a piece of glass in reflection mode (unless the piece has been too distorted by the firing process). Furthermore, during the scans the source-sensor geometry is kept constant in order to obtain an excellent uniformity of illumination. Even cheap scanners have spatial samples density in excess of 300 points per inch and light recording accuracy of 256 grey levels at least. The same quality is attainable at much higher costs with high resolution video cameras, equipped with orthoscopic lenses. Common video cameras cannot give high quality images because of the intrinsic limits of the TV standards. The main property of video cameras is the high speed of response, which is not so important for this type of application. Image scanners are integrated with computer software tools just like video cameras and, though slower, substantially offer the same readiness of use. A direct scan of thin flat glasses in the standard mode (reflection mode) gives good results which can be further improved by putting a reflecting film tight against the back surface of the glass, so as to enhance the signal by reflecting the forward scattering light. However, ancient glass often has a distorted surface (as in the observed case): the local curvatures are sometimes so high as to cause relevant special variations of illuminance and even cause reflections of the lamp light directly on the sensor. To avoid these effects, more than one light source was used for the illumination of the glasses from outside (transmission mode). This disposition allows better control of the angles of incidence of light, with sufficient uniformity of illumination.

After those laboratory tests the following passage was to put the scanner vertically on a wheeled rack mount, so as to easily place it tightly and gently against the panel. The images obtained recording the grisaille printing of the Madonna’s face and of the Child’s body were focused in spite of the fact that, for the relief due to the lead frame, the glasses were placed at a distance of about one centimetre from the scan plane. By performing a void scan without the panel, the incident light was checked so as not to have been directly seen by the sensor (i.e. by obtaining a completely dark image).

The next step will be to ascertain if it is better to extract the above mentioned details either from reflection mode images or from positive taken in the transmission mode, for instance after having placed a light diffusing screen in close contact with the glass from the side of the projectors.

As far as double glazing is concerned, the scanning method proves to be particularly practical because the results can be immediately compared with the originals by means of full scale laser printed transparencies. Common laser print resolutions are 600-800 dots per inch, which is sufficient to render good grey tone richness.

The aims of the research


Providing a useful tool for the historical-critical analysis of the stained glass windows by means of the images acquired with a computerized system and non-invasive method, integrated also with the images of the surviving grisaille strokes – as the complete reconstruction of the glasses in the face of the Madonna and those in the face and body of the Child. The tool also appears to show in detail the painting techniques used to lay the pigment.

b) Restoration

Making the lost paintings legible again, with no manual operation involved, by printing the recovered images on a suitable transparent plating material to be applied on the rear of the original glasses. The procedure should avoid the customary interventions to interpret the lost painting, made subjectively and often in difficult conditions, and thus perform the painting reintegration according to strict and objective methodological restoration criteria (4).

Other stained glasses have been recently examined and the results show that it is not always possible to obtain good results. The glasses subject to corrosion become ‘ opaque’ and that does not allow the printing recording. On the other hand, some times there is no printing left if the original heating process on the furnace was incorrect.

On some XVI century panels such as a Madonna and Child and St. Cecile , attributed to Innocenzo da Imola by Massimo Medica , that lost a large part of the painting – now being restored by Studio Fenice (5)- lighter printings, so difficult to record, have been found out, where positive image on the glass pieces are visible as in the two faces of the Madonna and Child and in that of St. Cecile, after the cleaning of the panel. That particular was noticed also by Marchini in the 70ies. He reported that in some cases, after treating the printings with graphite, the images reappeared in ‘positive’, and the fact seemed to be due to the ‘adhesion of the grisaille onto the glass’(6)

About what has been observed so far, the superficial lack of homogeneity in the printing (or ‘false ghost image’) of the Da Via Bargellini ‘s Madonna and Child, is not of the same nature noticed in the above mentioned cases.

A possible explanation is an insufficient heating exposure time in the furnace that could have prevented the formation of ‘undulations’ in the contact points of the grisaille and the glass, but anyway allowing the pigment to adhere on it for many years, being protected by any external agents. When the grisaille fell off ( it is supposed to be quite recent) the unpainted glass was already ‘opaque’ for its long exposition (that causes a light surface patina). Hence, the unexisting ‘false ghost image’ where once there was the grisaille.

The co-presence of both the types of image on the same window is to be found in the original executive process of the panels. It is also observed that only glasses of particular colours (‘carnicino’) give always ‘false ghost images’ on the windows by Innocenzo da Imola. With the green the image is usual, and with other colours there is a mixture of both. Thus, also the chemical composition of the glass is part of the different result as it influences on the adherence of the grisaille during the firing of the glass. This last theory coincides with Marchini’s but it should be studied deeper. (7)

Furthermore, it has been observed that those ghost images transmit an excessively weak signal to the computer system which at the moment cannot be detected by transmission also because its is disturbed by the several small surface scratches on the inner side of the tesseras (maybe due to incautious cleaning operations). Such an inconvenience can be read on the face of the Madonna scanner recording partially erasable applying a mathematic operator and with recording in reflection mode

Finally, it was noticed that the use of fixatives (like Paraloid B72)in the restorations, to make the falling grisaille adhere on the glass, alter the printing. That to highlight the importance of any slight trace that can give precious information should be maintained as far as possible.

Further aims for the future are:

  • defining the best illumination arrangement;

  • selecting the best image scanner and image processing software for the recording of the printing;

  • defining the procedure to erase from the recorded images the possible faults caused by small air inclusions;

  • selecting the materials suitable for the realization of plating glasses for restoration purposes.

It will be also interesting in the future to provide a selection of materials suitable to the realization of double glazing.

  1. During the restoration, the panel was cleaned with an aqueous solution (5%) of ammonium bicarbonate, and any paint left was fixed with a very dilute solution of Paraloid B72 in trichloroethane. Then, the panel structure was reinforced by spatula cementing and the broken pieces of glass were glued with epoxy resin.

  2. Once the pigment is fired on the glass into furnaces.

  3. cf. G. MARCHINI La restauration de la grisaille in ‘Verres Refract’, vol. 30, n. 1 January 1976.

  4. avoiding cold painting on the original.

  5. cf. R.P.FEYMAN, R.B. LEIGHTONM . SANDS The Feyman lectures on Physica, vol. 1, chapter 32, Addison-Wesley publishing Company, 1963.

  6. cf. G. Marchini, La restauration de la grisaille …

  7. In fact it is possible that other original faults take part in the formation of ‘false images’.



A.CASINI F. LOTTI L. STEFANI A. CORALLINI Un metodo di rilevazione e ricostruzione dei dipinti a grisaglia perduti. in "Le vetrate italiane: patrimonio da salvare – V – Conservare non restaurare", serie di seminari promosso dal Corpus Vitrearum Medii Aevi – Italia - Rendiconti dell’Istituto Lombardo Accademia di Scienze e Lettere, Milano vol. 128, (1994) 1997, pp. 211-230

ID - Un metodo di rilevazione e di ricostruzione del dipinto a grisaglia perduto applicato ad un manufatto dell’arte vetraia del XVI secolo in "Kermes" n.22 gennaio – aprile 1995, pp. 3-10

ID - F. CORTES PISANO Reconstruction of two 18th Century Rose Windows in the Cathedral of Girona, Spain in "News Letter" 48, maggio 2001, nuemro dedicato a "Le vitrail comme un tout", pp.