1. Introduction

First postulated by Briers [1], Laser Speckle Contrast Imaging (LSCI) is a well known technique that is primarily used in biomedical applications to study fluid flow and diffusion beneath thin transparent organic tissue such as skin and bone [2,3]. It has also recently been shown to be useful in examining targets which have no mechanical movement [4,5] or, on a microscopic level, are undergoing physical changes such as corrosion or absorption of water [6,7]. In this study we apply LSCI to the examination of simple cultural materials in order to resolve underdrawings that are hidden beneath a primary scattering layer on a macroscopic scale. This information is useful in determining the provenance of an item and assessing its material properties.

Speckle occurs when coherent light is either reflected from an optically rough surface or transmitted through a medium that has random refractive index fluctuations. By obtaining footage of speckle captured at a high frame rate in a short time period it is possible to look at the statistics of the intensity,I, as it varies over time at a specific point, $\left(i,j\right)$ , in the imaging plane. This information can then be used to spatially resolve details hidden beneath the initial scattering layer.

The used speckle contrast method most commonly used to view subsurface structure is the temporal contrast method [8]. It is a straightforward algorithm which examines the mean intensity, $〈I〉$ , and standard deviation,σ, over the entire time period,

The method described by Fujii [9] was used as an approach for real-time measurement of blood flow in the capillaries found on the surface of the hand. The final value, $F_{i,j}$ , is the sum of the contrast differences between consecutive frames

As an artwork is generally inhomogeneous on the surface it reasonable to assume that different regions of the surface will vary in temperature. This means that the time scales on which intensity changes occur vary across the sample. It becomes convenient to include in the calculations the differences between nonconsecutive frames. Arizaga’s [8] proposed method to deal with such variations is defined by the equation known as the weighted generalized difference (GD) equation,

In order to quantify the effectiveness of each averaging technique and the relationship between image contrast and temperature, the Michelson contrast formula,

2. Method

Figure 1 shows the experimental setup used to acquire images. A polarized 8 mW 633 nm HeNe laser illuminated the sample via a single mode fiber. The end of the fiber was oriented at 45° to the surface normal. The imaging system consisted of a monochrome 8-bit CCD camera (Sony, XCD-710) fitted with a 1:28, 50 mm video lens with an additional polaroid filter. Footage consisting of 50, 640 pixel × 480 pixel images, corresponding to an area of 16 mm × 12 mm, was collected at 30 frames per second. Data was saved in 8-bit TIFF format and later analysed within the program Interactive Data Language (IDL).

 

Fig. 1 Schematic diagram of experimental setup

Download Full Size | PDF

The general characteristics of each algorithm were obtained by analysing a simple monochrome image laser-printed onto the reverse of 80 gsm white paper. To quantitatively investigate the contrast produced by each of the algorithms using samples of area 30 mm × 30 mm consisting of 3 layers were constructed. White 220 gsm 2 mm mounting board was used as a base onto which one 30 mm × 15 mm rectangle was left blank and a simple 30 mm × 15 mm rectangle was drawn using one of five media: Conte Pierre Noir Pencil, Staedtler HB pencil, Artline 200 black pen, Artline 200 red pen, and charcoal. These rendered cards were then coated in Bostik Clag clear glue. A layer of either white 80 gsm paper, 220 gsm white card or 220 gsm gray card was attached. Some of these samples were also coated with two coats of Dulux Quick Dry matte white paint. Samples were prepared so that, in natural light, the underdrawing could not be seen with the unaided eye. Speckle footage was captured and analyzed with the three different techniques.

3. Results and discussion

3.1 Various samples and averaging techniques

The general features of each method are apparent in the images shown in Fig. 2 of a black image printed on the reverse side of 80 gsm white paper. Both the Temporal and Arizaga algorithms show significant darkening in the corners. Table 1 shows the corresponding contrast value between the white background and the medium used to draw the rectangle for the 3 layer samples.

 

Fig. 2 The three analysis techniques compared side by side (a) Original Image (b) Temporal (c) Fujii (d) Arizaga.

Download Full Size | PDF

Table 1. Contrast results for each analytical method applied to the various samples.

View Table

By inspection, contrast values below 0.1 did not reveal the underdrawing. It is clear that in all cases, the Arizaga method provided the greatest contrast and so was used for the rest of the experiments reported here.

The responses due to the actual media are of interest as it allows for the limitations of this method in the application to cultural materials examination to be explored. The contrast was greatest for the thinnest paper as it leads to maximum absorption by the underdrawing leading to a substantial speckle fluctuation. In the cases of the matte white paint and gray card the contrast was similar to that of the background value and, as expected, the black ink and charcoal led to the best contrast as these had higher absorption than the white ground surface.

3.2 Gray scale imaging

A gray scale laser printed image of four rectangles of varying degrees of gray was used to see if variations in shading could be resolved. The set up was as described above. Using the Arizaga algorithm qualitative detail resulting from variations in the depth of gray can be seen in Fig. 3 . This is useful as media in cultural materials are rarely, if ever, applied evenly during the construction process and so the subtleties and variations of the work can be observed to a certain degree. Intensity profiles taken across steps from darker to lighter regions for the original image and LSCI results are plotted in Fig. 3 (c). It can be seen that the spatial resolution of the LSCI images is clearly poorer than that of the original underdrawing with a step transition of approximately 5 pixels (original) and 50 pixels (LCSI of underdrawing).

 

Fig. 3 Arizaga result from the gray scale image that shows variations in the shading.

Download Full Size | PDF

3.3 Temperature dependence

An aluminium plate painted matte white with a resistor attached to the underside was used as the support surface for the mounting board - black ink - white card sample. A variable voltage was applied across the resistor, which heated both the plate and sample. The temperature was measured using a thermocouple placed on the edge of the sample. The voltage was slowly increased and images obtained when the temperature was stable. The images were then processed and the resulting contrast as a function of temperature shown in Fig. 4 .

 

Fig. 4 Contrast as a function of temperature.

Download Full Size | PDF

As can be seen in the plot in Fig. 4 there is a positive correlation between the contrast value and surface temperature. As the temperature of the surface increases so too do air currents above it, particularly where background media are present [11]. The Brownian motion of the air molecules leads to a decrease in speckle correlation resulting in a higher contrast between the underdrawing and background regions. It is apparent that the contrast saturates as the temperature of the sample exceeds 30°C. The frame rate of the camera is unable to adequately sample the faster fluctuations at the higher temperatures which results in an upper limit on the contrast value.

3.4 Application

As an example of the application of the technique to the examination of cultural materials, a book from the University of Melbourne collection, “The Life and Adventures of Robinson Crusoe” by Daniel Defoe (circa 1800, edition unknown), has been analysed. The pages in the book measure 20 cm × 13 cm and are 0.11 mm thick. Of particular interest in this study is a map that follows after the prologue. Due to repeated folding the page has failed down the length of the map and a recycled piece of paper with a table written on the visible side has been used to repair the break as seen in Fig. 5 .

 

Fig. 5 White light images of book sample.

Download Full Size | PDF

LSCI was applied to a 26 mm × 14 mm section of the scrap paper capturing 704 × 384 pixel² images at 30 frames per second and the data analysed with the Arizaga algorithm and lowpass filtered to remove noise.

While details aside from the table cannot be seen under white light, with the application of LSCI to the specimen we can reveal text that has been hidden by the surface. The most prominent being the word “Vinegar” (Fig. 6 ). This information is useful as many glue products used in bookbinding and repairs cannot be identified and removed so non-destructive techniques play a valuable part in analysis. This information can also be useful as an identifying feature of the book for cataloguing.

 

Fig. 6 (a) Reversed white light close u up of recycled paper (b) Reversed LSCI result revealing text beneath the table.

Download Full Size | PDF

4. Conclusion

LSCI is a straightforward technique requiring only relatively inexpensive components and has the ability to reveal subsurface drawings in paper. As the scattering and absorption properties of the surface materials increase, the spatial resolution and contrast of the underdrawing deteriorates until only noise is observable. This can be marginally improved by increasing the temperature of the surface which leads to enhancement in the LSCI image, but in the application of this method to heritage conservation this approach may be inappropriate. LSCI is useful in the practical application to repaired books by revealing identifying features that cannot be seen with the unaided eye. Future directions in this technique are the application of LSCI to parchment as the smoothness of the top surface has been demonstrated to give meaningful results. Here we have investigated the influence of several parameters on image quality, but other issues will also clearly play a role. While unable to image underdrawings beneath paint using visible light lasers, using an infrared light source may be of interest.