| NDT.net - May 2000, Vol. 5 No. 05 |
| TABLE OF CONTENTS |
In the second part the actual state of standardisation is discussed. A special test film has been developed and will be standardised world wide for the parameter evaluation of a digitisation system. Examples are given for two major applications of film digitisation : electronic archiving of films in connection with data bases and the printing of flaw catalogues. The reference catalogue ISO 5817 (assessment of weld imperfections, interpretation of arc-welded butt joints in steel) and the DGZfP catalogue D5 (Reference radiographs of castings) are produced by a digital film imager.
It can be foreseen, that improvements in this technology will considerable widen the application range during the next years.
2.1. Different Physical Principles
Film digitisation systems can be classified by the object's dimension that is sampled at each measurement:
| Principle | Scanner class |
| Point by point digitisation | Laser scanner |
| Line by line digitisation | CCD line scanner |
| Array digitisation | CCD camera |
Fig. 1 to 3 show the principles for each scanner class.
2.1.1. Point by Point Digitization
Fig.1 shows the principle of a laser scanner by the example of the LS85 scanner produced by Lumisys (U.S.A). The digitisation proceeds as follows:
The film is moved in front of an collection tube. A laser beam (wavelength about 680 nm, red) with a fixed diameter (ca. 50 µm) passes the film. The diffuse transmitted light through the film is integrated by the collection tube an registered by a photo multiplier (PMT) op top of the collection tube (not shown in fig.1). During the scan the folding mirror moves the laser beam along a horizontal line on the film. The film is moved with a speed of 75 lines/sec. The resulting voltage at the photo multiplier is proportional to the light intensity behind the film. After logarithmic amplification a digitisation with 12 bit yields grey values that are proportional to the optical density of the film.
The essential difference to other digitisation principles (cf. fig. 2 and 3) is the reversed optical way. The laser scanner illuminates with focused light and measures the diffuse light intensity behind the film. All other methods illuminate with diffuse light (the film is illuminated with a diffuser) and measure the light intensity that passes the film in one direction (camera objective or human eye in classical film inspection). The laser scanner principle has two main advantages:
|
| Fig 1: Principle of a laser scanner ( "LS 85", Lumisys, U.S.A.) |
2.1.2. Line by Line Digitization
Fig. 2 shows the principle of a CCD line scanner ("NDT-Scan II", DBA Systems, USA).
The film is illuminated by a light bar to which the light of a projector lamp is passed by a light guide. A Teflon plate ensures a diffuse film illumination. The illuminated line is projected by an objective on a CCD line detector. The film is moved under the light bar to scan the whole film area.
|
| Fig 2: Principle of a CCD line scanner ("NDT Scan II", DBA Systems) |
The combination of light bar, light guide, and projector lamp is the illumination of highest power (halogen lamp with 400 W) . Simpler and cheaper possibilities to illuminate the film are fluorescence lamps as used in DTP (Desk Top Publishing) and medical scanners. But these do not have the necessary brightness for a sufficient contrast sensitivity for optical densities up to D=4. For this reason such scanners cannot be used for NDT films.
Line by line digitisation is a good compromise between speed and influence of scattered radiation at high dynamical ranges (great differences between optical densities) that are typical for NDT films. Furthermore, modern CCD line arrays have operation modes for anti-blooming (avoid to transfer charges from saturated pixels to unsaturated neighbours) and integration time control. Thus, the scanner can be adjusted optimally to the optical density range of the film to digitise.
|
| Fig 3: Digitisation with CCD array camera |
There are also CMOS cameras available, which will give a logarithmic output signal relative to the input light intensity. So the digitised grey value will be proportional to the film density, and do not show to exponential behaviour as for CCD chips. This is a major advantage compared with CCD chips for the digitisation of radiographic films, but the signal-to-noise ratio for CMOS detectors is considerable lower than for CCD chips. The performance of actual CMOS chips is to low for NDT film digitisation, but this may be change in the next years.
Advantages:
Disadvantages:
2.2. Results of Investigation
Using a radiographic film of a test weld according to code EN 1435 (optical density range from 2.0 to 4.0) with many defect indications the image quality for each system can be determined visually. Fig. 5 to 8 show for each digitisation system:
Fig 4: The EPRI reference radiograph according to ASTM E 1936 and ASME Section V, Article 2, Mandatory Appendix VI and the CEN drafts
|
The size of the reference radiograph (14"x17") may be cut to custom fit a particular digitisation system (down to 8"x10") and still contain all of the necessary targets.
|
|
Fig 5: Scan results, profile plots, transfer curves and density contrast sensitivity of the Laser scanner LS85 SDR (Lumisys, U.S.A.), 50 µm pixel size, 12 bit logarithmic digitisation
| |
|
|
| Fig 8: Scan results and profile plots of the CCD camera "MEGA Plus" (Kodak), 70 µm pixel size, 10 bit digitisation; achieved density contrast sensitivity with an intensity proportional transfer curve | |
2.4. Comparison
The essential result of this comparison is the determination of the working and density ranges of each digitisation system. Within the densities of the working range the digitisation system has a density contrast sensitivity DDCS better than 0.02. A summary of all investigated systems is shown in Tab.1.
| Scanner Type | DD of the Working Range with Contrast Sensitivity DDCS < 0.02 | Density Range in which the Working Range is adjustable |
| Video Camera Sony AVC-D5 5122 Pixel, 8 Bit | 0.7 | adjustable in 0 ...4 |
| CCD Camera Kodak Mega Plus 10242 Pixel, 10 Bit | 1.0 | adjustable in 0 ... 4 |
| DTP Scanner Sharp JX-610 DIN A3, 12 Bit | 1.0 | adjustable in 0 ... 1.3 |
| CCD Scanner DBA NDT-Scan II 14"x17", 12 Bit | 1.9 | 7 fixed working ranges in 0 ... 4.0 |
| Medical CCD Scanner Scitron FD 60 14"x17", 15 Bit | 2.0 | fixed working range, 0 ... 2.0 |
| CCD Scanner Vidar Diagnostic Pro 14"x17", 12 Bit | 2.0 | fixed working range, 0 ... 2.0 |
| Laser Scanner CISE Alasca 14"x17", 12 Bit log | 3.0 | 2 fixed working ranges, 0 ... 3.0 and 1.0 ... 4.0 |
| Laser Scanner Lumisys LS 85 SDR 14"x17", 12 Bit log | 3.5 | 2 fixed working ranges, 0 ... 3.5 and 1.0 ... 4.5 |
| CCD Scanner DBA Image Master 14"x>17", 16 Bit | 4.0 (after merging) | adjustable in 0 ... 5.0 |
| Table 1: Ranking of investigated scanning systems according to the working range with a density contrast sensitivity better than 0.02 | ||
The results presented show that it is not possible to use simply scanners from DTP or medical applications to digitise NDT radiographs. Especially the necessary density contrast sensitivity up to optical densities of D>4 cannot be expected by these systems. These can only be obtained on scanners optimised for the NDT application.
Beside the requirements for maximum density and SNR, X-ray films require a very high spatial resolution. The limiting structure for very low X-ray energies is the grain size of the photo active silver based crystals, which is below 1 µm. This is particularly important for micro radiography. General NDT applications require X-ray energies between 50 and 12000 keV. In medicine, the application range is normally below 150 keV only. Due to this large energy range for NDT radiography, it was decided to reduce the requirements for spatial resolution to the unsharpness, which is caused by interaction of high energy X-rays with the screen film system. Measured functions provide unsharpness values between 30 and 800 µm, depending on the energy and the screen film system. Based on these measurements and the experience from the evaluation of different film digitisation systems (as shown in section 3), the following tables define the minimum requirements (see CEN draft, part 2 : minimum requirements).
Table 2 defines the minimum working range of the radiographic film digitisation system. In this working range, the digitizer shall provide an optical density contrast sensitivity DDcs which is DDcs £ 0.02. The minimum digital resolution is given for all devices converting the digital value proportional to the optical density. If the digital value is converted proportional to the light intensity, the digital resolution must be increased by at least 2 additional bits.
| Class DS | Class DB | Class DA | ||
| density range* DR | 0.5 - 4.5 | 0.5 - 4.0 | 0.5 - 3.5 | |
| digital resolution [bit] | ³ 12 | ³ 10 | ³ 10 | |
| density contrast sensitivity DDCS within DR | £0.02 | £0.02 | £0.02 | |
| Table 2: Minimum density range of the radiographic digitisation system with a minimum density contrast sensitivity | ||||
Table 3 specifies the minimum spatial resolution as a function of the X-ray energy.
| Energy | Class DS | Class DB | Class DA | |||
| [keV] | Pixel size [µm] | MTF 20 % [lp/mm] | Pixel size [µm] | MTF 20 % [lp/mm] | Pixel size [µm] | MTF 20 % [lp/mm] |
| £100 | 15 | 16.7 | 50 | 5 | 70 | 3.6 |
| £ 200 | 30 | 8.3 | 70 | 3,6 | 85 | 3 |
| £ 450 | 60 | 4.2 | 85 | 3 | 100 | 2.5 |
| Se-75, Ir-192 | 100 | 2.5 | 125 | 2 | 150 | 1.7 |
| Co-60 | 200 | 1.25 | 250 | 1 | 250 | 1 |
| Table 2: Proposed minimum spatial resolution of film digitisation systems | ||||||
On the basis of the image quality of film radiography and the state of the art of digitising systems, the working group has defined three quality classes; DA, DB and DS. The user may select the testing class based on the needs of the problem:
DS - the enhanced technique, which performs the digitisation with an insignificant reduction of signal-to-noise-ratio and spatial resolution,
Application field : digital archiving of films (digital storage)
DB -the enhanced technique, which permits some reduction of image quality,
Application field : digital analysis of films, films have to be archived,
DA -the basic technique, which permits some reduction of image quality and further reduced spatial resolution,
Application field : digital analysis of films, films have to be archived,
Due to the required international harmonisation, the standard reference film is taken over from ASTM for test and evaluation as well as for long term stability tests of digitisation systems.
We have studied the following application fields of film digitisation :
4.1. Application "Remote Inspection and Archiving of Turbine Blade Radiographs"
The German project "TRENDT" (Transmission and Archiving of Digital Radiographs - Application of Advanced Telecommunication in NDT) had been run successfully from 1992 to 1995 [5]. The project partners have been: DeTeBERKOM Berlin, providing the Berkom network; Du Pont, Bad Homburg, providing the ScanManager system for film digitisation as a basic system for the TRENDT application; SIEMENS KWU, Berlin, providing the test problem "remote inspection and archiving of turbine blade radiographs"; and BAM Berlin as a development partner to realise this application.
The aim of this project was twofold :
The realised user interface is shown in fig. 9.
Fig 9: User interface of the ScanManager station (TRENDT project) for archiving of the full film in reduced resolution (ROI1), annotated frames of the regions with full resolution (VIEW1) and the data base connection for film data and flaw types |
4.2. Latest developments
The problem on archiving huge numbers of digitised radiographs are the amount of data to be handled in a convenient way. So one film of 35x42 cm2 size digitised with 50 µm pixel size and 16 bit density resolution gives 112 MByte data. The lossless data compression used within the "Scan Manager" station for archiving on WORM media is limited to a factor of 2 caused by the noise inherent into the digital data. Lossy compression (e.g. DCT algorithm used by JPEG) is not usable for NDT applications because of the introduced artefacts. After decompression it must be ensured, that all indications originates from the digitised film. Artefacts caused by compression algorithms are not acceptable.
The solution of the TRENDT project was to reduce the spatial resolution of the complete film (overview) and to use the full resolution only in the interesting regions (ROI). But with the further development of huge storage media (RAID arrays of hard drives with several 100 GByte size, DVD-RAM's with 5 GByte and the today most used storage medium, CD-R with up to 800 MByte size) and the reducing costs for it this data size problem disappears during the next years.
An other recent development are the activities of the JPEG 2000 group [6] for standardisation of new image compression algorithms. It is expected to have a new industry standard available next year, based on the wavelet compression. This will have great influence to the image archiving community, because this state of the art algorithm allows to adjust image compression very accurately between lossless and lossy and support multi resolution too.
5.1. Scanning
Original films in the density range from D=0 up to D=4 have been scanned with the Laser scanner "LS85 SDR" from Lumisys. This scanner has a pixel size of 50 µm and an output depth of 12 Bit (up to 4095 shades of grey), the digital values are proportional to the optical density of the film.
Some films had densities higher than D>4.0. These high densities cannot be resolved by the Laser scanner "LS 85 SDR" with the necessary density contrast sensitivity. So a CCD scanner "Image Master" produced by DBA Systems was used. This scanner has a pixel size of 70 microns and reaches a contrast sensitivity DDCS in the density range from 0 up to 4.5 O.D. This wide dynamic range is realised by a special read out mode of the CCD. Each scan line is read out twice, but with an integration time differing by a factor of 100. The resulting two images are combined after scanning to one image (scan merging) covering this wide dynamic range. The digital scanner output is 16 bit proportional to the light intensity, but after the scan merging it will be transformed to 12 Bit proportional to the optical density of the film.
5.2. Rescaling of the digital data
After scanning the raw data with 12 bit depth are rescaled to 8 bit by a linear histogram optimisation. This step was necessary for a optimum printing output. The film printer accepts only 8 bit per pixel as maximum grey scale resolution. So the minimum and maximum data values of the scanned raw data corresponding to the maximum and minimum optical density of the scanned film were found and the difference between them was divided into 256 equally spaced intervals (8 bit). This rescaling was done interactively by visual presentation of each film on a high resolution monitor. This time consuming task could not be automated because of the great differences in the original films. The resulting 8 bit grey scale images of the complete catalogue was written to a CDROM in TIFF format for printing.
No other image processing (e.g. flaw enhancement, background suppression by convolution filtering) was used.
5.3. Printing
For printing the Scopix LR 5200 High Definition Imager produced by AGFA (see fig. 10) was used. This new Laser imager developed for Mammography illuminates a special laser film with a pixel size of 40 microns and a 16 bit grayscale resolution. A normal dark room development process follows this. The imager was loaded with an appropriate density curve giving a linear density range from D=1.5 up to D=3.5 for the 8 bit pixel values of the digital input image (see fig. 11). The saturation density in the background after printing is about D=3.8.
|
| ||
|
| Fig 11: Measured programmable transfer curves of the Laser imager LR 5200 |
The high background density and also the absence of any printing artefacts makes the image quality of this Laser imager superior to any other dry printing processes (Drystar by AGFA, DryView by 3M or Fuji). The printing resolution of these thermo sublimation printers is limited to 300 dpi (85 µm pixel size). This resolution is not sufficient to print all IQI's according EN 463 in original size. Also the maximum film density is limited to Dmax = 3.0. This limits the application of dry film imagers for NDT usage.
| © NDT.net - info@ndt.net | |Top| |