Document Imaging

The present invention relates to document imaging and in particular, but not exclusively, to apparatus and techniques for imaging books, especially rare books.

There is a growing trend for libraries, universities and other institutions to photograph rare books so as to preserve a record of their content in the event that damage or deterioration occurs to the original, and for web-site and facsimile production.

Typically the books are bound and it is necessary to support the book under a camera with its pages opened at the appropriate position. Techniques presently used are time consuming in that the camera must be adjusted and refocused after every page is turned in order to obtain a clear image. Typically it can take up to 10 minutes to image a single page. This means that it may take many weeks , even months , to photograph a whole volume . In some other systems that are faster the rare books can be in danger of damage and such systems are not appropriate. The present invention seeks to overcome or at least mitigate the above problem, and from a first aspect provides apparatus for positioning a page of a book for imaging, comprising a support for supporting the book with its pages open in a desired position, and a page positioning member for positioning a page in an imaging position, said page positioning member being movable from its page positioning position so as to permit a subsequent page of the book to be turned, the page positioning member then being returnable to its page

positioning position to allow the subsequent page to be positioned thereon.

Thus in accordance with the invention, it is not necessary to adjust and refocus a camera every time a new page is being imaged. A page positioning member, which is placeable at a fixed distance from an imaging device, receives a page for imaging. Once the page has been imaged, the page positioning member is movable from that position so as to allow a subsequent page of the book to be turned, most preferably while the book is still on its support. Once this has been done, the page positioning member may be returned to its imaging position, still at a fixed position relative to the imaging means and the new page positioned thereon. An image of the new page may be taken without any need to refocus or adjust the imaging means, since the positioning member, and thus the new page, is at the same spacing from the imaging means as previously. This allows a considerable saving of time over existing systems . Indeed it has been estimated that using apparatus in accordance with the invention, a new image can be taken every two minutes using a camera with a scanning back operating with an 80 second scan time - some five times quicker than with existing systems .

The page positioning member can take any convenient form so as to provide a page positioning surface.

Preferably, however, it comprises a plate. A plate can receive a page on its upper surface (the surface nearest the imaging means) and being relatively thin, fit easily between consecutive pages. Preferably the plate tapers in thickness in the direction towards the binding of the book to facilitate its insertion between the pages whilst being sufficiently stiff to ensure its flatness.

The page positioning member may be movable from its imaging position in any suitable manner. In a first embodiment it may be completely removed from the apparatus and then replaced after the book pages have been turned. Preferably, however, it remains attached to the apparatus to facilitate use of the apparatus.

Various ways of mounting the page positioning member are envisaged. For example, it may be translatable along one or more guides, for example linear guides. Preferably, however, it is pivotally mounted.

The book support preferably comprises a cradle having independently-adjustable means for supporting the front and the back of the book.

Most preferably, the support for the front of the book comprises a support surface mounted on a pivoted arm, such that the position of the surface relative to the arm can be adjusted in the direction normal to the support surface. The back of the book is supported on a similar support surface adjustable perpendicular to its support arm. This allows for proper positioning of the page being imaged as more and more pages are turned over.

The support surfaces may be adjusted with respect to their support arms manually or through other means such as jacks or the like. Preferably, the arms of the cradle can be adjusted relative to each other so as to accommodate a greater or lesser opening of a book. Preferably the cradle arms can accommodate an included angle of up to about 200° to allow for complete opening of the book, and reduce to about 90° to support books that cannot be opened fully.

The support surface for the book binding board behind the page that is being imaged need not be parallel

to the surface of the page positioning means, and it can be arranged in angle to it, for example typically 10° . This facilitates the access and the proper positioning of the page positioning means. The orientation of the apparatus relative to the vertical may be adjustable. This enables the camera lens axis to be set to more than one angle whilst maintaining the page positioning member perpendicular to the camera lens axis. There are two main modes for operation. In the first mode the camera lens axis is vertical or at a small angle to the vertical. This is for use when the book is opened through an angle of approximately 180°, for instance when a double-page spread is to be recorded, or when the book can be opened without damage through a relatively large opening angle such as 135° but less than 180°. In the second mode the camera lens axis is tilted through 45° or close to 45°. This mode is advantageous in that it places the document support cradle in a position in which it better supports thicker books and those which cannot be opened fully. In this second mode the whole book should preferably be opened through approximately 90° to minimise stresses on the book binding, and placed in the cradle with the front and back boards of the book each at approximately 45° to the horizontal . Thus the page being imaged is not itself horizontal but nevertheless is maintained perpendicular to the camera lens axis . Since the camera support and the book cradle and the page positioning member are all mounted on a common structural frame of the apparatus, the adjustment of the apparatus's vertical orientation maintains the correct positional relationship between the

support, the page positioning means and the imaging means irrespective of the imaging mode that is chosen.

The imaging means may be any suitable device. Typically, however, it comprises a high resolution digital camera.

A further problem that arises with document imaging is that, typically, the page being imaged must be located firmly against the page positioning means to avoid curling or warping and to ensure that it lies in the correct plane such that its image in the camera is in focus . The page positioning means may be a sheet of mild steel or other ferromagnetic material and this enables the page being imaged to be located in the correct plane for imaging by the use of one or more magnets, typically cylindrically-symmetrical pot magnets. However, this arrangement has the disadvantage that the magnet is opaque and may obscure and cast shadows upon the page being imaged.

Accordingly, the invention also seeks to provide an improved document location means, and from a second aspect provides a magnetic locator for a document, said locator comprising a magnet and a transparent locating member extending beyond the periphery of the magnet.

The transparent member which extends beyond the body of the magnet can be used to hold the document against the ferromagnetic page positioning means. This member will not obscure the margins of a page and will, therefore, allow a fuller image to be captured.

Preferably the transparent member is a plate-like member. Typically, thickness of the member is 2 mm or less, more preferably 1 mm or less.

Preferably the edge of the transparent member is chamfered. This has been found to cast less shadow on the underlying page and thereby improve the image quality. In order to reduce the magnetic gap between the magnet and the underlying ferromagnetic page positioning means, and thereby maintain a strong locating force, the magnet is preferably recessed into the upper surface of the transparent member . The transparent member may be made of any suitable material, for example glass. Preferably, however, the transparent member is made from a plastics material, such as an acrylic material e.g. polyethylmethacrylate (Perspex"") or a polycarbonate. This reduces the risk of fracture of the material during use (possibly damaging the document being imaged) and also facilitates fabrication of shapes more complex both in plan and in edge profile.

A yet further problem which arises during imaging of books is that distortion of an image occurs in the region between facing pages . This is a particular problem with tightly bound books where a significant gutter may be formed between facing pages. In many ancient or historical volumes there are marginal annotations or features such as illuminations that extend into the gutter region and which may be of just as much interest as the main document text .

Generally distortion of an image will take place in two perpendicular planes, namely perpendicular to the binding edge (i.e. laterally across the page) and parallel to the binding edge (i.e. up and down the page) .

Distortion perpendicular to the binding edge is caused by two effects, namely foreshortening due to the curvature of the gutter and a reduction in magnification (compared to the remainder of the page which lies in the imaging plane) due to the increased distance between the page in the gutter and the imaging means .

Parallel to the binding, the distortion results merely from the effect of the loss in magnification due to increased distance from the imaging means . The present invention in another aspect seeks to overcome this problem and provides a method for correcting distortion of an image of a gutter region between facing pages of a book, said method comprising the steps of positioning a reference pattern in said gutter region; imaging said gutter region with said reference pattern in position to obtain a distorted image of the reference pattern; calculating the correction that would have to be applied to the distorted image to remove the distortion; imaging a page without the reference pattern; and applying the calculated correction to the image of that page .

Thus in accordance with this aspect of the invention the distortion of a reference pattern is used to calculate a correction for correcting that distortion. This correction can then be applied to the images of other pages to remove (or at least reduce) the distortion in those images .

In a simple embodiment, the reference pattern can comprise a series of lines extending parallel to the binding line. If the spacing between the lines is known, the distortion due to the curvature of the gutter region can be determined and a suitable correction calculated

that would produce an image with the lines at the correct spacing.

It is possible to calculate the profile of the gutter from the distortion of the reference pattern and knowledge of the distance between the page and the lens of the imaging system. This allows the distance of the various points in the gutter from the imaging means to be calculated, allowing the effect of the loss in magnification due to the greater spacing of the gutter from the imaging means to be calculated and corrected. This correction is applied in addition to the curvature related correction to allow correction of the image in the direction parallel to the binding.

The reference pattern need not extend across the full width of the page. It is important for best results that it extends as far as possible into the binding in the gutter and extends outwardly across the entire width of the gutter.

The reference pattern need not extend across the full height of the page. In a simple embodiment, it can extend over just a proportion of the page height, preferably in the mid height region of the page. The system is suitably configured to check in the appropriate region of the page for the presence of the reference pattern.

With such an arrangement, it is assumed for the purposes of applying the correction to the image, that the gutter profile is constant over the height of the page. If greater accuracy is needed, then the pattern could extend over a greater area of the gutter, or providing a number of patterns at different positions. This may be preferable since a pattern extending over a

large area of the gutter may not be able to follow the profile of the gutter without itself being distorted.

The reference pattern is provided on a flexible carrier such that the carrier can conform to the gutter profile. Preferably the carrier is flaccid to improve the ability of the carrier to conform to the profile.

Preferably the carrier is a plastics sheet material.

Preferably the carrier is opaque such that the reference pattern is not obscured by the underlying page. It is not necessary to calculate a separate correction for each page being imaged. The curvature of a gutter will remain generally constant over a number of turned pages, so that a single correction can be applied to a number of pages, a new correction then calculated and so on.

The invention also extends in broad terms to a method for correcting distortion of an image of a gutter region between facing pages of a book, said method comprising the steps of comparing an image of a reference pattern positioned in the gutter region with the original reference pattern; and calculating a correction that would have to be applied to the image to remove any distortion in the image.

The present invention also extends to a system for implementing the methods defined above.

From a further aspect the invention therefore provides a system for use a reference pattern positionable in the gutter region of a book; processing means for processing a captured image of said reference pattern in said gutter; said processing means comparing the captured image data against reference data for the reference pattern and calculating a correction to be

applied to the captured image data to bring it into conformity with the reference data.

The system may employ standard data processing equipment suitably programmed to effect the method. Accordingly the invention also extends to software for carrying out the method of the invention. More particularly the invention provides a computer software product adapted when run on a data processing machine, to carry out the steps of: comparing an image of a reference pattern positioned in the gutter region of a book with the original reference pattern; and calculating a correction that would have to be applied to the image to remove distortion in the image.

The process of removing the distortion in the image may be undertaken as soon as both an image without the reference pattern and a corresponding image with the reference pattern have been recorded, but to avoid delaying the capture of subsequent images the process is best performed off line. Some preferred embodiments of the invention will now be described by way of example only with reference to the accompanying drawings in which:

Figure 1 shows an apparatus in accordance with the invention in a first orientation; Figure 2 shows the apparatus of Figure 1 in a second orientation;

Figure 3 shows a sectional view of a first embodiment of a page locator in accordance with the invention; Figure 4 shows a plan view of the embodiment of

Figure 1 ;

Figure 5 shows a sectional view of a second embodiment of page locator;

Figure 6 shows an open book with a graticule applied in accordance with the invention; Figure 7 shows a plan view of Figure 6 ;

Figure 8 shows an image taken of the arrangement of Figure 7 ;

Figure 9 shows a corrected image in accordance with the invention; and Figure 10 shows the graticule of Figures 6 to 8 in detail.

With reference firstly to Figures 1 and 2, a document imaging apparatus in accordance with the invention comprises a frame 4 indicated schematically, an imaging device 6, a book support cradle 8 and an page positioning plate 10.

The imaging device 6 is preferably a digital camera which is connected to an image processor (not shown) . It is supported on a support arm 12 which is vertically adjustably mounted on a column 14 of the frame 4.

The page positioning plate 10 is mounted below the imaging device 6. It is pivotally mounted about a hinge 16 whereby it may be pivoted vertically in the sense of Figure 1 away from its imaging position. The plate 10 tapers in thickness in the direction away from the hinge 16.

The support cradle 8 is arranged below the page positioning plate 10. The cradle 8 comprises first and second arms 18, 20 which are independently pivotable about an axis which is fixed in position relative to the frame 4 and the hinge 16. The included angle between the arms can typically be from 90° to 200°. Each cradle arm

18, 20 mounts a jack mechanism 22, 24 having a surface 26, 28 for supporting the front and back of a book 30. The arm 20 may have a larger angular travel than the arm 18. For example the arm 20 may move through 90° (such that the surface 28 can travel between the limit positions shown in dotted lines in Figure 1) while the arm 18 may travel through a smaller angular range.

As can be seen from Figure 1, the surfaces 26, 28 can support the front and back of a book 30 while a page 32 of the book 30 is positioned on the upper surface 36 of the page positioning plate 10 for imaging. The page 32 will be located on the plate 10 (which is typically of steel) by locating magnets which will be described further below. To image a page 32, the camera 6 can be moved relative to the page positioning plate 10 along the column 14 until the image of the main part of the page 32 is in sharp focus. At that point a image of the page 32 is taken and captured electronically. Once the image is taken, the page 32 is folded back and the page positioning plate 10 rotated anticlockwise about the hinge 16 to the vertical position shown in dotted lines 10'. This allows the next page 32 for imaging to be turned over while the book 30 is still on the cradle 8, the page positioning plate 10 then being returned to its original imaging position and the next page 32 then folded down over and positioned on the plate 10. Since the page 32 is in good contact with the page positioning plate 10, the distance between consecutive pages being imaged and the camera does not change, meaning that the camera will still be in sharp focus, avoiding the need for any adjustments to the camera.

This process can then be repeated for as many pages as need to be imaged.

It will be appreciated that as respective pages 32 of the book are turned, then the number of pages 32 supported on the arm 18 under the page positioning plate 10 will decrease. In order to compensate for this, and keep the page being imaged as close as possible to the imaging plane, thus minimising gutter effects, the jack 22 may be raised from time to time so as to reduce the spacing between the support surface 26 and the page positioning plate 10.

With fragile books, it is desirable to be able to support the documents with opening angles not greatly exceeding 90° and with the front and back boards having similar angles to the horizontal, typically 45°. To accommodate this, the apparatus 2 may be arranged as shown in Figure 2. This is essentially the same as the arrangement shown in Figure 1, except that the frame 4 has been rotated through approximately 45°. Suitable structure should be provided to support the frame 4 in both orientations .

It should be noted that to facilitate imaging, it is not necessary to have the support surface 26 parallel to the page positioning plate 10. Indeed, positioning of a page 32 is easier if the support surface 26 is at a slight angle e.g. 10° to the plane of the page positioning plate 10.

As mentioned above, the page 32 is located in position on the page positioning plate 10 by one or more a magnetic locators. A first embodiment of a locator in accordance with the invention is shown in section in Figure 3.

The locator 40 comprises a magnetic body 42 which is mounted to a transparent plate 44 by a screw 46 or some other suitable attachment mechanism, such as adhesive. As will be seen, the transparent plate 44 extends beyond the periphery of the magnetic body 42 such that the projecting part 48 can be positioned over a marginal region 50 of a page 32 being imaged. In this way, the magnet 42 will still act to clamp the page 32 in position, but it will not interfere with the image of the marginal region 50.

In order to increase the clamping force afforded by the magnet 42, it may be mounted in a recess 52 in the upper surface 54 of the transparent member 44, as shown in Figure 5. It will also be noted from Figure 5 that the edge 56 of the transparent member 44 is chamfered. This reduces the shadow formed on the underlying page, thereby improving the image. It is preferred that this chamfer does not have any sharp edge for this reason. Turning now to Figures 6 to 10, these Figures illustrate image correction in accordance with the invention.

As can be seen from Figure 6, when a book 30 is opened, a gutter 62 is formed in the binding region between adjacent pages 32. The effect of this is to distort the image taken of the page 32. This is because while the main part 64 of a page 32 being imaged will be positioned on the page positioning plate 10, and thus at a fixed distance from the imaging device 6, the gutter region 62 curves away from the part 64. This leads to two distortion effects. The first is due to the curvature of the gutter region 62 and the second is due

to the change in magnification of the image due to the greater distance the gutter region 32 is positioned away from the imaging device 6.

In order to overcome this problem a graticule 66 is used in the calculation of a correction to apply to the image. In the embodiment shown, the graticule 66 comprises a sheet 68 of flaccid, opaque plastic material which is able easily to conform to the profile of the gutter 62. The graticule 66 comprises a printed reference pattern, in this embodiment a series of lines 70 which are spaced from one another at a regular, known interval. Its outer extent is marked with a wide bar 72 (Figure 10) . In order to correct distortion, an image of a page 32 is first taken without the graticule 66 in position. This image will suffer distortion due to the effects discussed above.

The graticule 66 is then positioned on the page 32. As shown in Figure 7, the graticule 66 is preferably placed in a central region of the page 32, although this is not essential and it could be placed in any suitable position on the page 32. The graticule need not extend across the full width of a page, provided that it extends as far as possible into the binding and extends outwards across all the gutter region 62 into the flat region 64.

When the graticule 66 is in position, a further image is taken of the page 32 and the image fed to an image processor, which can be a standard data processing machine suitably programmed. The image taken is arranged to cover the entire page that is to be corrected, together with a small gutter part 62 ' of the facing page

32'. The part 62' of the facing page 32' is sufficiently- wide that the image space which it occupies is sufficient to accommodate the corrected version of the image of the required page 32. This enables the corrected image to have the same number of pixels as the as taken image.

The as taken image is illustrated in Figure 8. It will be seen in the gutter region 32, the page suffers both distortion in the direction A, parallel to the binding of the book 30, and also in the direction B, perpendicular to the binding, as discussed above. The distortion perpendicular to the binding can be corrected by analysis of the image of the graticule 66. Since the spacing between the lines 70 in the graticule 66 is known, the image processor can manipulate the image data so as effectively to stretch the taken image in the gutter region 62 so as to provide a constant graticule spacing as per the original. In particular, the distortion of the image in the B direction is fitted with a function to provide smoothing of the line position errors and permit interpolation between the lines. This will allow mapping of distorted image points into a distortion corrected image.

The part of the image in which the lines 70 are constantly spaced (the part of the flat part 64 of the page 32) is not affected by this correction.

The correction can be effected by redistributing image-information data between the pixels in the relevant part of the image. The correction is stored for use on pages to be imaged. Correction of the image of the page 32 introduces a distortion to the imaged gutter region of the facing page 32', but this is not significant for the image of the

page in question. If required, the portion of region of the facing page 32 ' can be removed electronically, for example by over writing the relevant pixels with a suitably chosen background colour or by cropping the image to remove the facing page .

Now the distortion in the direction A parallel to the binding must also be corrected by the image processor. As mentioned above, this is due to a magnification error due to the increase in distance from the imaging device to the page surface as the gutter is approached. The shape of the gutter can be calculated from the function fitted to the distorted image discussed above in the production of the first distortion correction. Once the page profile is determined, the magnification error at any distance from the gutter can be determined and the page stretched accordingly. Thus the image processor can calculate the distance from the imaging device to the various points in the gutter, calculate the magnification error and calculate a suitable correction, effectively stretching the page in the direction A. The amount of stretching will vary in dependence on the position in the B direction within the gutter.

It is assumed in the distortion correction calculations that the page curvature is always in one direction, i.e. convex towards the imaging device.

In the preferred embodiment, individual distorted image pixel data (e.g. the RGB or other value at pixels) is mapped into a corrected image. As the extent of distortion will not generally be a multiple of the pixel spacing, the corrected image pixel RGB or other values will generally be a weighted combination of. the values

from the distorted image pixels close to the position corresponding to that of the pixel in the distorted image. As an example, a bicubic interpolation can be used for this purpose. It will be appreciated that it is not necessary to image a graticule on every page being imaged, since the page profile for a number of successive pages will not vary greatly. Accordingly, a series of pages can be imaged and then a page imaged with and without and the graticule, then a series without the graticule then a further page with a graticule and so on throughout the book.

To implement the correction automatically, file names of page images with superposed graticules can be flagged such that they are recognisable both as graticule images and by their place in the sequence of pages. The correction program then proceeds in turn through the pages to be corrected and for each page identifies the graticule image with the nearest sequence number and then uses that graticule image as the basis of the correction. The above process assumes that the page profile is generally constant over the height of the page . If a more accurate determination is required, graticules may be provided at more than one location along the height of the page and suitable corrections applied depending on the position of the particular image section along the height of the page .

It is preferred in this circumstance to use a plurality of individual graticules rather than one graticule which extends over the entire surface of a page since a number of individual graticules can more easily conform to the page surface .

It will also be appreciated that other forms of graticule patterns can be used, for example, a reticular pattern. What is important is that the particular graticule pattern is known such that a correction can be applied to the actual image produced so as to correct back to the original image .

While it is preferred to place a physical graticule on the page to gauge distortion, it may be possible to project a pattern onto the page for the same purpose. In such a process, the pattern can be used to profile the page gutter from which a correction in both of the perpendicular directions (parallel to and perpendicular to the book binding) can be calculated.

It will be understood that while the description above is in the context of imaging bound volumes, the apparatus of the invention can be used to image individual pages suitably positioned on the page positioning plate.