Title of Invention

"A PROCESS AND DEVICE FOR THE AUTOMATIC INSPECTION OF THE PRINT QUALITY OF IMAGES"

Abstract

The invention relates to an electronic device for automatically inspecting the print quality on deformable objects by means of an elastic model. The model is firstly produced by capturing with an electronic camera (CCD for example) the images of a set of sheets whose print quality is regarded as acceptable; the images are stored so as to produce a first reference image, together with the relevant densitometric tolerance limits. This reference image is thereafter divided into a multitude of sub-images by superimposing a grid with very small mesh cells. During inspection, the distances between the nodes of the grid are measured on the image to be inspected: this therefore produces an elastic modification of the model, which is such as to make the distances between the nodes the same as in the image to be inspected. The image to be inspected is thus verified with respect to the modified reference (model) by using any of the standard inspection techniques.

Full Text

The present invention relates to a process and device for the automatic inspection of the print quality of images FIELD OF THE INVENTION The present invention relates to a device for the very careful checking of the print quality on deformable materials, such as sheets of paper, plastic or rubber. More precisely, the invention constitutes a device which uses a process for modifying, in real time during inspection, the reference models customarily used in automatic checking of print quality so as to correctly inspect, even in cases where the deformation of the sheets produces distortions such that the printed shapes turn out very differently from the model, but nevertheless are still acceptable to the human eye. In fact, all the systems known hitherto use rigid models (undeformable), and hence, in order to cater for the deformations of sheets (paper, plastic, rubber), are compelled to considerably relax the tolerances so as to reduce the risk of false detections. In particular, this produces a very often unacceptable reduction in the carefulness of inspection. Several processes for judging the quality of print are generally known: a few examples are given in the list of references. Although several authors have proposed a great variety of setups, almost all the solutions are based on the same fundamental approach, which can be aptly summarized as follows. A set (Training Set, TS) of one or more samples of objects (sheets, etc.) with a "good" quality of print is used to "train" the inspection system using the reference model, and (or) the relevant printing tolerances in terms of densitometric measurements. In general the process consists in capturing, by means of an electronic scanning system, image(s) of sample(s) of the TS and of the construction of a "reference model" (sometimes referred to as a "golden template") which may be the average of the TS images (or some desirable tansformation of them) . In addition to the densitometric reference value (that is to say the value in the reference image), for each pixel of the analyzed image, a pair of limit values is calculated (for example, too dark TD and too light TL). Several techniques have been proposed to extract these limits: for example some authors use the minimum (over TS) of the densities of the pixel as TD, and its maximum as TL; others use the gradient of the image; others the standard deviation; etc. In any event, the reference model is a description of the printing tolerances which associates the densitometric limits TD and TL with each pixel of the image. These descriptions are "rigid", that is to say there is no possibility of catering for deformations, which produce a relative displacement of the printed structure. Therefore, in all the previous solutions, despite the introduction of a few cunning processes for relaxing the tolerances of the thresholds (TD and TL) , the deformations of the carrier (paper, plastic, etc.) are the main source of the detection of "false defects", that is to say of prints which do not have defects as far as a human inspector is concerned, and which notwithstanding this are rejected by the system. In addition, an increase of this kind in the tolerances causes the inspection to become rather coarse and inaccurate, consequently reducing the standard of quality. SUMMARY OF THE INVENTION According to the present invention, the objects (sheets of paper, plastic, rubber, etc.) to be inspected are analyzed optically by well known optoelectronic means, such as for example a CCD camera (linear or matrix, black and white or color) , with the desired resolution so as to produce electronic images of the printed sheets. An electronic image consists of a discrete set of density values, generally organized as a rectangular matrix. Each element of the matrix (pixel) is a measure of the intensity of the light reflected by the corresponding part of the image. These density values are often digitized into 256 gray levels (for example zero corresponds to black, 255 to white). In the case of color images, the description generally consists of three matrices corresponding to the red, green and blue components of the light reflected, by each pixel of the image. In what follows, the term "image" is always used instead of "electronic image", both in the case of black and white images (a single matrix of density values), and in the case of color images (three matrices). As in several solutions from other authors, a training set (TS) must first be defined, consisting of one or more images of the "properly printed" sheets (that is to say, those with no unacceptable defects), which will be used to construct a "model" of good sheets. The system must therefore proceed according to the following steps: extract from TS an image to be used as a reference (which could be one of the TS images, their average, or any suitable transformation, such as for example the image of the edges of the printed shapes on the sheets) divide the reference image into a large number of sub-images, for example by superimposing a grid (regular or otherwise) as shown in Plate I/4b) choose at each mesh cell of the grid a very characterized structure of the printed shape (see Plate II/4), whose position will be used to measure the deformations of the carrier. In what follows we shall refer to the positions of the said structures as the nodes of the model. the structures may vary from certain very sophisticated topological characteristics, to other rather simpler ones, such as the maximum of the gradient of the image inside the mesh cell. A few techniques for automatically extracting the characteristic structures, and hence the nodes of the model, will be illustrated in the "description of the preferred embodiments" define for each node a deformation threshold as the maximum acceptable displacement of the node from its position in the reference image. Lastly, it is possible to construct (from the TS images) for each pixel of the model, the densitometric thresholds (for example, too dark TD and too light TL) according to any of the techniques specially used for this purpose (max-min; standard deviation; variation of the gradient; etc.), but before they are calculated, the system carries out the following processing on each TS image: measure the displacement of each node of the TS image with respect to the corresponding node in the reference image produce an elastic deformation of the TS image such that all its nodes have the same position in the reference image. This step can be undertaken with the desired accuracy by using any of the well-known image deformation algorithms, such as for example the one referred to as "2-pass mesh warping" (see reference 1). The device which is the subject of the present invention therefore constructs the model on the TS images suitably deformed so that the nodes have the same position as in the reference image. During inspection the images to be evaluated in order to check their print quality are firstly processed in the same way as the images of the TS, and thereafter compared with the limits of the model (thresholds) as in other approaches. Such a process ensures that deformations which are smaller than the deformation thresholds defined elsewhere will be corrected, so as to guarantee very careful inspection, while excessively deformed sheets will be rejected as defective. Finally, it is obvious that the same result can be obtained by deforming the model (together with its limits, for example the thresholds TD and TL) instead of correcting the images to be inspected. To meet the aforementioned objectives the invention provides for a process for automatically judging the print quality of images printed on a deformable carrier, such as sheets of rubber, plastic or paper, by using an optoelectronic image capture device and an image processing system which measures the deformation of the carrier by superimposing a suitable grid on the image to be inspected, and thereafter by measuring the displacement of the nodes of the grid with respect to their position in an image regarded as an undeformed reference image, which, before comparing them with the thresholds, deforms the images to be inspected so that their nodes have the same position as those of the reference image. The invention also provides for a device for the automatic inspection of the print quality of images printed on a deformable carrier, such as sheets of rubber, plastic or paper, using an optoelectronic image capture device and an image processing system which measures the deformation of the carrier by superimposing a suitable grid on the image to be inspected, and thereafter by measuring the displacement of the nodes of the grid with respect to their position in an image regarded as an undeformed reference image, which, before comparing them with the thresholds, deforms the images to be inspected so that their nodes have the same position as those of the reference image. BRIEF DESCRIPTION OF THE DRAWINGS Plate 1/4 shows at a) the character "A" as an example of a printed shape, and at b) the superposition of a 3 x 3 grid. Plate II/4 shows at a) five characteristic structures of the shape in Plate 1/4, and at b) the associating of five nodes (from Kl to K5) with these structures. Plate III/4 shows at a) an example of the deformation of the character "A" of Plate II/4, and at b) the positions of the nodes in the deformed shape, highlighting the corresponding displacement for each node. More precisely, if Ki [i e (1,5)] is the position of the ith node in the original image, and K'i its position in the deformed image, it is found that in our example all the displacements Δxi and Δyi are zero, except for Δx2 and Δx5. Plate IV/4 shows a typical arrangement of the inspection system described in the text. DESCRIPTION OF THE PREFERRED EMBODIMENTS As an example of the present invention, one of the preferred embodiments will be described hereinafter on the basis of the drawings. Plate IV/4 shows an arrangement of the inspection system in which a linear CCD camera 1, with its lenses 2 and its illumination system 3, is used to capture the images of the sheets 4 to be inspected while they rotate around the cylinder 5. The lines scanned by the camera are stored in sequence in a first buffer (memory) circuit of the image processing subsystem 6 to produce an (electronic) image of each sheet. The image processing subsystem 6, which could be based either on special hardware or on programmable computers, such as DSP (Digital Signal Processors), very fast PCs, etc., carries out different operations during the model construction phase and the inspection phase. During the model construction phase: it captures the images of the sheets of the TS and stores them in an appropriate memory it extracts the reference image from the TS (either automatically, or with the aid of the operator, by means of the console of the Operator Interface 7) it superimposes a grid onto the reference image. The number of rows and columns of the grid can either be predefined or entered by the operator by means of the Operator Interface 7 it identifies the position of a node in each mesh cell of the grid, as the coordinates of the pixel or the quantity q given by (Equation Removed) at its maximum over the mesh cell In equation [1] I(P) is the value of the electronic image at the position of the pixel P and A is a very small square surface (a few pixels) whose center is at P. By maximizing the product of the sum over A of the absolute values of the partial derivatives we ensure that the node is a structure whose vertical and horizontal position are easily detectable. Thereafter, each image of the TS is processed by the image processing subsystem 6 according to the following steps: identification of the displacement Ax,Ay of each node of the TS image with respect to the corresponding node in the reference image. In this embodiment the operation is performed using the method of maximum correlation: a small rectangular portion So of the reference image, with its center on the coordinates x0,yo of the node, is compared with a portion Si, with the same dimensions, whose center is displaced step by step onto each position (pixel) of the TS image so as to find the position x1,y1 at which the correlation coefficient has its maximum. The displacement is then given by Ax = xi - x0 and Ay = yi - yo deformation of the TS image so as to make its nodes have the same position as in the reference image. In this embodiment, the operation is performed by using the already-mentioned algorithm referred to as "2-pass mesh warping" (see reference 1) calculation over the deformed images of the TS of the average Avg(P) and of the standard deviation Sgm(P) of each pixel of the image. During the inspection phase, according to the present embodiment of the invention, the image processing subsystem 6: firstly effects on each image to be checked, captured by the camera 1, the same deformations used during the model construction phase. The image to be checked will therefore be deformed in such a way that its nodes have the same positions as in the reference image thereafter, calculates the difference A(P) between the value I(P) of each pixel P of the image to be checked and the corresponding value of the average Avg(P) finally, rejects as defective any sheets which are overly deformed (that is to say, sheets for which the displacement of at least one node is larger than the already-defined thresholds). It also rejects sheets for which Δ(P) > KSgm(P) over at least M pixels inside an area with center P and radius R. The parameters K, M and R can be chosen by the operator so as to define the carefulness of inspection (by means of the Operator Interface 7). The other preferred embodiments include: a) the substituting of a matrix camera for the linear camera. In such a case, the illumination subsystem will have to use flash devices synchronized with the image frequency of the camera, to ensure correct, capture of the image. b) the use of the average of the images of the TS as a reference image c) the manual selection of nodes d) the use of a multitude of small masks (templates), (each with its center on a node) to search for the displacement of the nodes with a technique such as that referred to as "best superposition" (or best matching) e) the approach already mentioned in which, instead of deforming the image to be inspected in such a way as to position its nodes as in the model (reference image), the model is deformed in such a way that its nodes have the same position as in the image to be inspected f) the use of a different definition for the thresholds from that of proportionality to the standard deviation. Claims 1. A process for automatically judging the print quality of images printed on a deformable carrier, such as sheets of rubber, plastic or paper, by using an optoelectronic image capture device and an image processing system which measures the deformation of the carrier by superimposing a suitable grid on the image to be inspected, and thereafter by measuring the displacement of the nodes of the grid with respect to their position in an image regarded as an undeformed reference image, which, before comparing them with the thresholds, deforms the images to be inspected so that their nodes have the same position as those of the reference image. 2. The process as claimed in claim 1, in which the reference image is one of the images of the TS (Training Set, defined in the text). 3. The process as claimed in claim 1, in which the undeformed reference image is the average of the images of the TS. 4. The process as claimed in any of the preceding claims, in which the nodes are positioned manually by the operator on characteristic structures of the printed shape. 5. The process as claimed in claims 1 to 3, in which the nodes are positioned automatically on the maximum inside each mesh cell of the quantity defined by equation [1]. 6. The process as claimed in claims 1 to 3, in which the nodes are identified automatically by extracting inside each mesh cell a characteristic structure (characteristic either from the topological or geometrical point of view, such as in particular contours, edges, angles, segments, holes, etc.) by means of any technique for extracting well-known characteristic structures. 7. The process as claimed in any of the preceding claims, in which the deformation of the images is obtained by means of the already-mentioned algorithm referred to as "2-pass mesh warping". 8. The process as claimed in any of the preceding claims, in which the deformation of the images is calculated by approximation according to techniques other than the "2-pass mesh warping" algorithm, including, but not limited to, elastic deformations, bilinear interpolation or interpolation of any degree, and rigid translation of the mesh cell. 9. The device as claimed in any of the preceding claims, in which, instead of deforming the images to be inspected so as to obtain the same positions of the nodes as in the reference image, the model (reference image) is deformed so as to obtain the same positions of the nodes as in the image to be inspected. 10. The process as claimed in any of the preceding claims, in which the nodes are searched for over regions of the sheet in a manner different from the superposition of the grid, including random searching. 11. A device for the automatic inspection of the print quality of images printed on a deformable carrier, such as sheets of rubber, plastic or paper, using an optoelectronic image capture device and an image processing system which measures the deformation of the carrier by superimposing a suitable grid on the image to be inspected, and thereafter by measuring the displacement of the nodes of the grid with respect to their position in an image regarded as an undeformed reference image, which, before comparing them with the thresholds, deforms the images to be inspected so that their nodes have the same position as those of the reference image. 12. The device as claimed in claim 11, in which the reference image is one of the images of the TS (Training Set, defined in the text). 13. The device as claimed in claim 11, in which the undeformed reference image is the average of the images of the TS. 14. The device as claimed in claims 11 to 13 in which the nodes are positioned manually (by means or the operator interface) by the operator on characteristic structures of the printed shape. 15. The device as claimed in claims 11 to 13, in which the nodes are positioned automatically on the maximum inside each mesh cell of the quantity defined by equation [1]. 16. The device as claimed in claims 11 to 13, in which the nodes are identified automatically by extracting inside each mesh cell a characteristic structure (characteristic either from the topological or geometrical point of view, such as in particular contours, edges, angles, segments, holes, etc.) by means of any technique for extracting well-known characteristic structures. 17. The device as claimed in claims 11 to 16, in which the deformation of the images is obtained by means of the already-mentined algorithm referred to as "2-pass mesh warping". 18. The device as claimed in claims 11 and 17, in which the deformation of the images is calculated by approximation according to techniques other than the "2 pass mesh warping" algorithm, including, but not limited to, elastic deformations, bilinear interpolation or interpolation of any degree, and rigid translation of the mesh cell. 19. The device as claimed in claims 11 to 18, in which, instead of deforming the images to be inspected so as to obtain the same positions of the nodes as in the reference image, the model (reference image) is deformed so as to obtain the same positions of the nodes as in the image to be inspected. 20. The device as claimed in claims 11 to 19 in which the nodes are searched for over regions of the sheet in a manner different from the superposition of the grid, including random searching. 21. A process for automatically judging the print quality of images substantially as herein described with reference to and as illustrated by the accompanying drawings. 22. A device substantially as herein described with reference to and as illustrated by the accompanying drawings.

Documents:

3-del-2000-abstract.pdf

3-del-2000-claims.pdf

3-DEL-2000-Correspondence Others-(27-12-2011).pdf

3-del-2000-correspondence-others.pdf

3-del-2000-correspondence-po.pdf

3-del-2000-description (complete).pdf

3-del-2000-drawings.pdf

3-del-2000-form-1.pdf

3-del-2000-form-13.pdf

3-del-2000-form-19.pdf

3-del-2000-form-2.pdf

3-del-2000-form-3.pdf

3-DEL-2000-GPA-(27-12-2011).pdf

3-del-2000-gpa.pdf

3-del-2000-petition-137.pdf