WO1992009002A1

WO1992009002A1 - Optical scanning apparatus and method

Info

Publication number: WO1992009002A1
Application number: PCT/GB1991/001733
Authority: WO
Inventors: Robert Massen
Original assignee: British Technology Group Ltd
Priority date: 1990-11-07
Filing date: 1991-10-07
Publication date: 1992-05-29
Also published as: DE4035368A1

Abstract

A scanning system adapted to produce enhanced resolution in an optical scanning device includes means (21, 24) to generate a composite light spot by the superposition of a plurality of light (21, 22) spots of differing size and frequency, means (26) to direct the composite light spot towards a corresponding plurality of frequency selective photoreceivers (28, 29), the photoreceivers being so constructed or selected that each has a peak sensitivity response to only one of the light frequencies and signal processing means (32) to generate from a weighted (30, 31) sum of the response of said photoreceivers a response which corresponds to the scanning with a two-dimensional transfer function of predetermined shape.

Description

Optical scanning apparatus and method This Invention relates to methods of enhancing the resolution of resolution of images produced by scanning and, in particular, to image scanners and printers employing such methods.

The resolution which may be obtained with optical scanning and printing systems such as laser scanners, laser plotters and printers is inversely proportional to the size of the projected light spot. This Invention describes a technique, which despite a relatively large light spot size gives a high spatial resolution and a system transfer function which may be tailored for specific characteristics, such as high-pass filtering. At least two light spots with differing sizes, light frequencies and centre positions are superimposed on the surface of interest the reflected light from which 1s detected by a number of frequency selective light detectors. The output video signal of the system is generated from the bipolar weighted sum of each individual detector's electrical signal. Through selection of the spot size, centring and weighting a large number of useful system transfer functions especially those with high-pass characteristics, may be Implemented. In contrast to two-dimensional electronic signal filtering this spatial filtering technique Is independent of geometrical distortion introduced by the individual detectors and scanning mechanisms.

The same method may also be used as a printing method. Here at least two light spots with different sizes and frequencies are projected superimposed upon each other on the surface of interest. The photosensitive surface to be illuminated is made of a mixture of at least two fine structured frequency selective photosensitive materials with opposing transfer characteristics.

Optical scanners which acquire an image in a predefined, usually point by point and Hne by Hne, manner are used in many industrial and commercial applications: 1. For the optical inspection of surfaces in the steel, glass, textile, wood, paper, coating and film industries the, usually quickly, moving surfaces are scanned in a Unewise fashion by means of a laser scanner. The light scattered from the surface is converted to an electrical video signal by means of detectors, such as photo ultipliers or photodiodes. From this electrical signal the quality of the surface may be determined: scratches, holes, grooves etc. display themselves as Intensity variations in the video signal and may be detected, for example, by means of threshold monitoring techniques.

2. Several scanners sample surfaces by scanning an image of the surface of interest past a light sensitive detector by means of a rotating polygon mirror, a galvano etric mirror scanner or similar. This type of scanner use an external uniform light source for illuminating the surface to be scanned.

3. Laser printers and film printers also seek to use a light spot, focused as finely as possible, to scan across a light sensitive surface ⁽photosensitive film, photosensitive semiconductor coating etc.) to generate a desired image 1n a point by point fashion.

All these scanners and printers have a common feature. To attain the maximum possible resolution a light spot must be focused as sharply as possible. The more sharply the light spot of a surface scanner is focused, the smaller the defect which may be resolved. The sharper the spot of a laser printer, the sharper the generated image.

The generation of such fine light spots is limited by physical/optical laws. The requirements of optical elements such as beam expanders or, spatial filters, the demands on coherence, the frequency and mode stability of the laser and thus the costs of the system increase disproportionately to the spot diameter reduction.

The generated light spots usually have a symmetrical or asymmetrical two-dimensional Gaussian Intensity profile. The optical signal which is received by the photodetector is the convolution of this profile with the structure of the sampled surface. The wider the intensity profile the more smearing of light/dark edge structures 1s caused leading to a loss of image contrast in neighbouring fine structures and poor sharpness in printed Images.

An Improvement in the resolution and contrast may be obtained by processing the video signal with known Image processing methods. These methods, such as high pass filtering and edge enhancement, are most usually Implemented by two-dimensional electronic signal processors. Such methods can, however, only be implemented with difficulty with existing laser scanners. Laser scanners have, along each Hne, an uneven scanning speed and polygon mirrors giving rise to Hne flutter leading to geometrically distorted and unstable images. The use of electronic two-dimensional filters and convolvers, especially when used in 'real time' on the digitised video signal require a geometrically faithful scanning of the surface with very Httle geometrical distortion. Such geometrical stable scanners are very difficult and costly to realise, especially for the scanning of large surfaces at high speed. The use of electronic digital two-dimensional filters, despite their known advantages in giving Improvements to contrast and edge definition, is therefore restricted.

This Invention describes a method and a technique wherewith point for point scanners and printers may give a high resolution and contrast despite having a large Hght spot and/or a strongly geometric distortion, to yield Images comparable to those obtained with much smaller light seats or by additional digital two-di ensional signal processors. This is obtained by the superposition of at least two light spots having differing sizes, Hght frequencies and spot centres on the surface of interest. The reflected light is captured by the same number of frequency selective light detectors the electrical signals of which are processed by weighted summation. According to the present Invention there is provided a scanning system having a transfer function adapted to produce enhanced resolution in an optical scanning device which uses a point sampling method including generating means to generate a composite light spot by the superposition of a plurality of light spots of differing size and frequency, means to direct the composite light spot towards a corresponding plurality of frequency selective photoreceivers, the photoreceivers being so constructed or selected that each has a peak sensitivity response to only one of the light frequencies and signal processing means to generate from a weighted sum of the response of said photoreceivers a response which corresponds to the scanning with a two-dimensional transfer function of predetermined shape.

The invention will now be particularly described with reference to the accompanying drawings, in which:

Figure 1 shows the sampling of a black/white object by means of a laser with an ideal (upper) and a typical (lower)

Hght spot size and the resulting video signal.

Figure 2 shows the relationship between the Hght spot Intensity profile, the system impulse response/aperture function and the spatial frequency transfer function.

Figure 3 shows a typical contrast improvement of high frequency structures obtained with a bipolar two-dimensional aperture function with an appropriate high-pass spatial frequency transfer function.

Figure 4 shows the implementation of such a bipolar aperture function through the superposition of two Hght spots with differing sizes and light frequencies. The surface scattered light is detected by two frequency selective photodetectors, converted to two electrical signals which are and summed with bipolar weighting coefficients.

Figure 5 shows an application of this invention for the printing of films where also two light spots with differing sizes and frequencies are superimposed on the film. The photosensitive coating on the film is comprised of a mixture of two fine grained frequency selective materials with opposing transfer characteristics. Figure 6 shows a schematic diagram of a system which is one possible implementation of the technique described 1n Figure 4. As an example of one possible use of this invention a laser scanner for surface inspection will be described. The surface to be inspected is, for the purposes of this example, taken to be smooth with dark scratches and dark surface defects upon it which are to be detected. As shown in Figure la, an ideal (theoretically infinitely small⁾ spot is scanned over the surface under test. Figure lb shows the video signal obtained from such ideal sampling. Such infinitely sharply focused light spots are, of course, due to physical laws not realisable in practice. A laser spot has a symmetrical or asymmetrical Gaussian intensity profile with a finite size. The size of the aperture function, I(x)., of this profile causes a smearing of the steep light/dark edges of the surface defects in the video signal, S(t). Due to this smearing closely neighbouring fi e structure are only resolvable with poor contrast.

In many surface scanners, an unwanted directional dependency of the intensity of the scattered Hght received by the photodetectors is often present (the so-called shading error). This gives rise to an unwanted slope in the video signal base-line, which may give rise to difficulty in the recognition of defects when fixed amplitude thresholds are used. High-pass filtering of the video signal can restore a correct base-line

The behaviour of such a scanning system may be explained in terms of electrical and optical system theory. The Gaussian form of the two-dimensional intensity profile, I(x,y), defines the optical aperture function of the Hght spot. This, 1- jrn, defines the Impulse response of the system, that . , the response of the system to the Dirac pulse.

It 1s known from system theory that the frequency transfer function and impulse response function are a Fourier transform pair. Fourier transformation of the two-dimensional aperture function I⁽x,y⁾ gives the two-dimensional spatial frequency transfer function H(u,v), of the optical systems light spot. Spatial frequency is defined as the number of (sinusoidal) light/dark transitions per unit length. Units of spatial frequency are often given in, for example, line pairs per mm. The Fourier transformation of a Gaussian function is also a Gaussian function. The spatial frequency transfer function of a scanning light spot with a Gaussian intensity profile is consequently a low pass function. The resolution limit of the scanner is determined by that corner frequency, (ug.vg), at which the value of H(ug.vg) 1s still acceptable. The normal definition of this corner frequency is the -3dB value of H(u,v).

As the Hght intensity Kx.y⁾, may only take positive values, 1t is only possible to generate scanning spots with a low-pass spatial frequency transfer function. The only possibility to realise a better resolution and better contrast for edges and small structures is to increase the corner frequency (ug.vg), by reducing the light spot size (x0,y0). This always implies a great technical effort and higher costs in realising the necessary expansion and focusing optics and places stringent demands on the coherence of the laser used etc.

A higher resolution may also be attained by an electronic high-pass filtering of the video signal to compensate for the low-pass characteristics of the gaussian spot. Figure 3 shows a well-known two-dimensional high-pass filter used for edge detection and recognition in digital image processing. The bipolar form of the impulse response h(x,y) has a 'Mexican Hat¹ form. By means of Fourier transformation one obtains the spatial frequency transfer function, H(u,v), whose high pass characteristic gives a selective sharpening of higher spatial frequencies. It is known how to make such improvements to a video signal by software on digitised and stored Image data or by means of special hardware convolvers on the digitised data stream. With such digital filters, user defined impulse function (including bipolar functions) may be implemented. The use of such digital filters however demands a geometrically exact image else the necessary neighbour relationships between the pixels within the convolution kernel are not met. This geometrical accuracy and stability 1s only rarely met with most laser scanners due to the non-constant scanning speed along a line, to the non-constant Hne separation along the traverse Erection caused by mirror bearing play (the so-called Pyramid eϊror) and similar non-ideal behaviour of the mechanical and optical parts of the scanner.

It is very costly to produce laser scanners and printers with the high geometric stability required for a good working two-dimensional digital filtering of the captured video signals able to compensate the low-pass behaviour of the spot.

The Invention described herein permits the design of optical scanners and printers with bipolar high-pass system transfer functions. They rely on the superpositioning of several light spots with different wavelengths, each having a unipolar system Impulse response and on the bipolar summation of the separately frequency selective detected light signals.

This technique is explained 1n more detail in Figure 4. Figure 4a shows the superpositioning of two symmetrical Gaussian

Hght spots with differing radii, xl, x2 and differing frequencies (wavelengths) f1 , f2. Figure 4b shows that a bipolar weighted subtraction for a bipolar weighted addition of both profiles, Il(x.y) and I2(x,y) gives rise to a resulting intensity profile I3(x,y) showing the desired 'Mexican Hat' form. Light intensities as purely positive quantities may not be optically subtracted, at least not in the case of non-coherent Hght. This invention shows that this subtraction may be performed electronically for this purpose Instead of one optical detector two frequency selective ones must to used. This may be implemented, for example, by means of an optical band-pass filter 1n front of each photo detector. The centre frequencies of each filter are set to correspond to the emission characteristic of each of the light sources. The electrical signals of each of these detectors are multiplied by a weighting factor, wl and w2, and summed. When one of the weights is negative one has the desired subtraction and thus the desired high-pass system response characteristic.

By careful selection of the radius of the superimposed Hght spots, the centring of the centre of the spots and the summing weights a multiplicity of useful isotropic and anisotropic system apertures with corresponding two-dimensional high-pass spatial frequency transfer functions may be implemented. The number of superimposed light spots may be larger than just two to give a greater freedom in approximating a required system transfer function.

The two-dimensional filter function obtained by this Invention Is largely independent of geometrical errors in the scanners. This Is because the transfer function solution relies solely on the optical characteristics of the scanning light spots and not on the digital processing of a geometrically faithfully reproduced electrica'l video signal.

If the Hght spot sizes and radii are chosen such that the positive and negative sections of the system aperture are equal the whole system responds as a pure two-dimensional high pass filter which is useful for the above mentioned shading correction.

If the centres of gravity of the two superimposed Hght spots are for example offset in the line direction, one obtains a high-pass filter characteristic only in this direction giving rise to an accentuation of vertical defects. In total a very large number of two-dimensional bipolar .iso- and anisotropic system apertures may be syntheslsed by careful selection of the three parameters spot radii, spot centre offset and weighting coefficients. A further advantage is that this technique requires no storage of the scanned image and thus may be readily implemented in real time synchronous to scanning. It may be implemented as a purely analogue circuit or digitally after digitisation of the video signal with minimal effort. It may also be implemented purely optically when using a weighted summation of different polarised or phased light spots.

A second example of application of this Invention is the contrast and resolution improvement of laser printers or film printers. This example is shown in Figure 5. A light spot 21 focused as finely as possible on a photosensitive material, 22, such as photofilm or selenium semiconductor coated drum. As shown in Figure 5a the size of the Gaussian intensity profile gives rise to unsharpness in the image points produced. In practice it is assumed that the photosensitive granules are much smaller than the laser spot. Due to the illumination energy the transparency or darkness of all granules or the electrical characteristics of a photo-sensitive semiconductor material increase as described by the slope of the characteristic response curve of the material, T(I), 24.

The Implementation of this invention for this application is shown in Figure 5a. At least two Hght spots 25, having differing sizes, frequencies and, if desired, differing centre positions 26, are superimposed on a finely grained photosensitive mixture 27 composed of two frequency-selective, granule components and whose response characteristics, 28 29 are, at least in part, opposing.

The bipolar weighted summation is implemented by the sign and slope of the characteristic response curve of each component of the mixture. The summation occurs in the eye of the observer who is Incapable of resolving individual grains but whose eye has a local integrating action.

This invention may also be used by scanners which do not project a light spot on the object surface and gather and convert the scattered light but rather use a polygon mirror wheel or galvanometric mirror scanner to scan a photodetector over the surface of interest. In this case the desired bipolar system for achieving sharp images is achieved by arranging the photoreceiver 1n the form of a number of individual photosensitive annular elements whose form is that of a slice, or contour line, through the desired two-dimensional system aperture.

The individual annular photodetector zones are Individually connected and the total output signal is formed from the weighted sum of the electrical signal from each of these annular elements. The weighting coefficients can be implemented by analogue resistors or by digital multipliers but also by the size of each annular element.

One possible circuit implementation, but by no means the only one, is shown in Figure 6. Two lasers, 21 and 22 with differing frequencies, fl and f2 are superimposed by means of a partially reflecting mirror, 23. The light spot sizes are individually set by means of the optical elements, 24 and 25. With the aid of a rotating polygon mirror, 26, the superimposed light spots are scanned over the surface of interest, 33. Two photoreceivers, 28 and 29, each of which responds to only one of the laser frequencies, gather the scattered Hght and convert it to an electric signal. Both signals are multiplied, for example by means of, an analogue multiplier, with the weights wl and w2 and then subtracted by the summing amplifier 32. Obviously the weighted summation may also be performed on the digitised signals with digital circuits that are well known to practitioners of the art. The output video signal S(t) then appears as if it were obtained using a two-dimensional bipolar aperture from a spot with a Mexican Hat intensity profile

The setting of the free parameters, such as light spot size, intensity, centring and weighting may be understood by a competent person from this description. A desired two-dimensional spatial frequency transfer function with, for example a high-pass characteristic for edge enhancement, may be implemented. By application of the inverse Fourier transformation to the required system transfer function one obtains the associated two-dimensional impulse response or system aperture. From this, by means of known approximation techniques, such as least squares fitting, the above mentioned free parameters may be determined.

The applications described herein are typical examples, the technique is not limited to just surface Inspection with laser scanners and laser printers and film printers as here discussed.

Claims

PATENT CLAIMS 1. A scanning system having a transfer function adapted to produce enhanced resolution in an optical scanning device which uses a point sampling method characterised in that it includes 5 generating means to generate a composite light spot by the superposition of a plurality of light spots of differing size and frequency, means to direct the composite light spot towards a corresponding plurality of frequency selective photoreceivers, the photoreceivers being so constructed or selected that each

10 has a peak sensitivity response to only one of the Hght frequencies and signal processing means to generate from a weighted sum of the response of said photoreceivers a response which corresponds to the scanning with a two-dimensional transfer function of predetermined shape.

152. A scanning system having a transfer function adapted to produce enhanced resolution in an optical scanning device which uses a point sampling method as claimed in claim 1 characterised in that it includes generating means to generate a sampling composite Hght spot by the superposition of a plurality of

20 Hght spots of differing size and frequency, the composite light spot being scattered by a scanned surface and received by a corresponding plurality of frequency selective photo receivers, the photoreceivers being so constructed or selected that each has a peak sensitivity response to only one of the light

25 frequencies and signal processing means to generate from a weighted sum of the electrical outputs of said photoreceivers a video signal which corresponds to the scanning with a two-dimensional transfer function of predetermined shape.

3. A scanning system adapted to enhance the resolution of 30 optical scanning devices which use a point sampling method as claimed in claim 2 characterised in that it includes guidance means to ensure that at least two of said plurality of light spots are not coincident.

4. A scanning system as claimed in any one of the preceding 35 claims characterised in that it includes an annular arrangement of photodetector elements and adder means to perform a summation of the electrical signals of these elements to form an output video signal, the size and the shape of the annular elements being chosen such as to correspond to iso-he1ght slices of a 5 required two-dimensional impulse response of the scanning system and the weights of the summation circuit corresponding to the signed amplitude of the associated slice of the impulse response.

5. A scanning system as claimed in claim 4 characterised in that said adder means 1s adapted to perform a bipolar-weighted

10 summation of said electrical signals.

6. A scanning system as claimed 1n claim 1 characterised in that said photoreceivers comprise a mixture of at least two frequency selective photosensitive materials with, at least 1n part, opposing response characteristics.

157. A scanning system as claimed in claim 1characterised in that it comprises at least two light sources with differing frequencies, an optical superi poser a single superimposed spot may be generated, mirror scanning means to deflect said spot over a surface the Hght of which is scattered and collected 0 with the aid of at least two light receivers, each of which has a peak response to one of the light frequencies and the electrical output signs of each receiver being multiplied with a stored weighting coefficient.