DIGITAL BIOMETRICS, INC. v. IDENTIX, INC

PLAGER, Circuit Judge.

Digital Biometrics, Inc. ("DBI") brought suit against Identix and Randall Fowler for patent infringement, both literal and under the doctrine of equivalents, in the United States District Court for the Northern District of California. The district court granted Identix's motion for summary judgment of non-infringement. See Digital Biometrics, Inc. v. Identix, Inc., No. C 95-01808 CW, 1996 WL 784567 (N.D.Cal. Dec. 16, 1996). DBI appeals. We affirm the district court's judgment of non-infringement.

A. BACKGROUND 1.

The present suit involves U.S. Pat. No. 4,933,976 (the `976 patent), entitled "System for Generating Rolled Fingerprint Images." As the title implies, the patented invention relates to a system (and method) for capturing, storing, and displaying fingerprint images. Unlike conventional systems, which use paper and ink, this one uses a computer controlled imaging and retrieval system. By digitally representing a fingerprint image, the system can automate fingerprint storage, retrieval, and most importantly, searching.

The inventors of the `976 invention were not the first to create a digital fingerprint system. Several such prior-art systems are described in the "Background of the Invention" section of the patent. The problem with those systems, according to the patent, was that they failed to represent accurately the fingerprint image due to the problems associated with capturing a curved surface of the fingerprint on a flat, two-dimensional surface. Accordingly, the stated advantage of the present invention is that it produces "fewer discontinuities and artifacts" in the final image. Col. 2, ll. 22-23.

The system described in the written description of the patent includes a microprocessor coupled to random access memory (RAM) and read only memory (ROM). The microprocessor is further coupled to a terminal, such as a standard keyboard, that allows an operator to input data. A video monitor and printer are also included in the system to allow the operator to view and print, respectively, a fingerprint image generated by the system.

The method of generating the fingerprint image is at the heart of this patent and at the heart of this dispute. It is therefore necessary to describe precisely the imaging sequence performed by the system. First, the system generates an analog representation of a fingerprint image, referred to in the written description as "frames of video signals." Col. 3, ll. 52-54. In the preferred embodiment, that analog image is generated by a video camera that captures an image of a fingerprint presented on an "image propagating surface." In the preferred embodiment, this surface is a prism or platen, upon which a subject rolls a finger. The camera captures the entire image of the image propagating surface of the prism. Col. 3, ll. 49-51. Because of the arcuate shape of the finger, however, only a portion of the finger is actually in contact with the prism at any time. Thus, the camera, in effect, takes "snap-shots" of the image propagating surface of the prism at discrete intervals of time as the finger rolls across the surface.

The analog output of the video camera is fed to a "frame digitizer" that converts the analog output to a digital format. In the words of the written description, "[d]igitizer 24 produces two-dimensional arrays of digital pixel values PVn, m/ representative of the intensity of fingerprint images at corresponding discrete pixel locations PLn, m/." Col. 3, line 67 — col. 4, line 2. In the preferred embodiment, the frame digitizer is an 8-bit analog-to-digital converter. Col. 4, ll. 17-20. The digitizer operates under the control of the microprocessor. The output of the digitizer is read by the microprocessor and stored in RAM. The data structures produced by this process are repeatedly referred to as "image arrays." See, e.g., Col. 4, ll. 2-4, 16; col. 5, ll. 9, 27, 38, 47, 50, 53, 65; col. 6, ll. 12, 47, 43-54. The "image arrays" are referred to as "IAL, IAC and IAR" in the written description, as shown in Figs. 3A, 3B, and 3C, respectively, for the example illustrated therein. Each image array includes "N horizontal rows and M vertical rows." Col. 4, ll. 6-7. In one preferred embodiment, N is equal to 480 and M is equal to 512. This size is apparently a direct function of the size of the frame signal from which the image array is derived. As shown in the drawings, for purposes of illustration in the application, N is equal to 20 and M is equal to 28. The exact point in the process at which the optical image becomes an "array," however, is one of the disputed issues in the case.

Once the first image array is captured and stored in memory, the system determines the extent to which the intensity of each pixel varies from that of an average intensity. Col. 4, ll. 51-55. These resulting "variances" are then used to define an "active area" within the image array. Generally, the higher the variance for a particular pixel the greater the probability that that pixel actually represents that portion of the finger currently in contact with the image propagating surface. An active area is therefore a sub-set of the image array in which there exists data characteristic of an actual fingerprint image. In the preferred embodiment, the active area is a rectangular array. The left and right edges or boundaries of the active area are determined by comparing the variances in a particular row with a pre-determined threshold value. The microprocessor compares each variance value to the threshold value. The left edge of the active area is set equal to the column of the image array in which the first variance value exceeds the threshold value. Col. 5, ll. 50-67. The microprocessor then sequentially processes the variances in the particular row in question until it identifies the last variance value in the particular row that is greater than the threshold value. The right edge is then set equal to the last so identified column. Col. 6, ll. 4-9. The upper and lower boundaries of the active area are assumed, in the preferred embodiment, to be equal to row 1 and row N, respectively.

The pictures in the specification are particularly helpful in illustrating the actual operation. Figure 3A, shown below, illustrates a first image ("OIL") taken from the platen as the finger is rolled from left to right.

On the subject of drawings, we share Identix's concerns over DBI's use of its drawings in their brief. Usually drawings are quite helpful, particularly in cases involving complex technology such as this, and are to be encouraged. Figs. 4-6 and 8 of DBI's brief are a hybrid combination of drawings from the patent and additional figures from an unknown source. Because the drawings incorporate figures from the actual patent they have an air of legitimacy. Figure 4 does have an accompanying footnote that admits that it is "not specifically illustrated in the patent," but no such disclaimer appears on Figs. 5-6 or 8. Moreover, Fig. 4 fails to indicate from whence it came. Only upon a search of the record did we discover that similar, but noticeably distinct, versions of these figures were included in DBI's expert's declaration in opposition to Identix's motion for summary judgment. Thus, those figures are arguably misleading and not in an inconsequential way. For example, the versions that appear in the expert declaration are labeled "storage" while those same boxes are labeled "RAM" in their present brief. In addition, the "data characteristic" of a "slice" is labeled differently between the two. The need for "storage" and the definition of "slice data" are the two primary issues in the case. Even though Identix decries the inclusion of these drawings it has not requested that they be stricken. Accordingly, we consider them as we would any argument, based on their intrinsic merit as revealed by the record.

The corresponding image array is shown in Fig. 4A. For the image array IAL shown in Fig. 4A, the corresponding active area is shown by the broken-line AAIAL.

Once the active area is defined, the image data within the image array circumscribed by the active area boundaries, and only that data, are copied into a "composite array," which is used to store the final fingerprint image. Col. 6, ll. 14-20. This step is described in the written description as "copying the first `slice' or active area such as AAIAL into composite array CA." Col. 6, ll. 37-8. The composite array will previously have been initialized so that all of the values therein are cleared. Figure 5A, shown below, illustrates the results of copying active area AAIAL into the composite array.

The above described sequence is substantially repeated for each subsequent video frame. As the fingerprint is rolled across the image propagating surface, subsequent frames will contain a different portion of the fingerprint image. The current video frame is digitized in the above-described manner and stored as the next image array. That image array is then analyzed by the microprocessor to determine its active area. The active area for that image array is then merged into the composite array, as described further below.

The result of the image capturing and active area defining steps for the second video frame is illustrated in Figs. 3B and 4B, respectively, as shown below. The active area for this second image array is labeled "AAIAC" in Fig. 4B.

The active area of this second image array AAIAC is then combined with the active area of the first image array using a "mathematical function." Overlap between the current active area AAIAC and the previously stored active area in the composite array AAIAL is critical to the effective operation of the system. In fact, if the microprocessor fails to detect such an overlap, it activates an alarm to indicate to the operator that a "portion of the fingerprint is . . . lost." Col. 7, ll. 24-30. The existence on an overlap is easily determined in the present invention by comparing the left edge of the current active area with the right edge of the previous active area, as now stored in the composite array. In the preferred embodiment, the system requires a minimum overlap of eight pixels. In the illustrations shown, however, a different minimum is assumed.

Once the microprocessor determines that there is in fact an overlap, it merges or combines the current active area with the composite array using the mathematical function. In order to carry out this merging process, the microprocessor compares each pixel in the current active area with the corresponding pixel in the composite array. If the pixel in the current active area is lighter than the corresponding pixel in the composite array, the composite array pixel is left unaltered. Col. 8, ll. 16-33. If, however, the current active area pixel is darker than the composite array pixel, the two are averaged and the result is stored in the composite array pixel location. Col. 8, ll. 34-48. By retaining the darker composite array pixels and averaging in the darker current active area pixels, the system produces "a smoother transition between two adjacent slices resulting in a composite array CA more accurately representative of the rolled fingerprint image." Col. 7, ll. 45-48.

2.

The inventors filed an application for a patent on January 25, 1988. A preliminary amendment was filed on September 1, 1988, canceling claims 1-2 and adding claims 3-25. In an Office Action dated March 30, 1989, the Examiner at the United States Patent and Trademark Office ("PTO") rejected all of the claims. Claims 3-4, 19-22 and 25 were rejected under 35 U.S.C. § 102(b) in view of Ruell (German Patent No. 3423886). Claims 5-14, 16, 23 and 34 were rejected under 35 U.S.C. § 103 solely in view of Ruell. Claims 3-18, 21-23 and 24 were also rejected under 35 U.S.C. § 112, at 2, for failing to particularly point out and distinctly claim the applicants' invention. The Examiner did indicate, however, that claims 15, 17 and 18 were allowable if rewritten in independent form and amended to overcome the § 112, at 2 rejections. The applicants submitted an amendment on September 28, 1989 along with accompanying remarks. That amendment canceled claims 4, 20 and 21 and added new claims 26-38. In response to this amendment, the pending claims were ultimately allowed.

Identix is a competitor of DBI's in the market for computerized fingerprinting systems. Randall Fowler is the President of Identix. Identix manufactures two products — the TP-600 and the TP-900. Like the patented invention, the TP-600 is a microprocessor-based system. The TP-600 is comprised of an imaging system that is coupled to an analog-to-digital ("A/D") converter. The imaging system is substantially similar to that described in the patent. Identix does not contend otherwise. The TP-600 imaging system produces an analog signal that is representative of a fingerprint image using a charge-coupled-device ("CCD"). The signal level of the analog signal corresponding to a particular pixel is a function of the image intensity at that pixel. The analog signal has lower values for ridge information and higher values for valley information, similar to the darker and lighter representations generated when ink is used. The analog signal produced by the CCD is converted to a digital format by the A/D converter. The A/D converter produces a stream of digital values that correspond to and are synchronized with the stream of analog data produced by the CCD. Based on the timing of the analog signal, the system knows precisely to which image pixel in the CCD's two-dimensional matrix the analog signal corresponds.

The TP-600 also includes a memory system. This memory system is used to store the fingerprint image produced by the imaging system. Each digital data value produced by the A/D converter is stored in a location in the memory system corresponding to its pixel location in the CCD. Before a digital value is stored in its corresponding memory location, however, the current digital value produced by the CCD is compared to the existing value in that memory location. While the current digital value is being compared, it is temporarily held in a register. For purposes of the present appeal it is enough to say that the system uses a mathematical function to determine what value to store in a particular memory location based on the value of the current digital value and the existing digital value.

Arguably the TP-600 uses a "logical" as compared with a "mathematical" function to decide which of the two values to store into the memory location. We consider this a distinction without a difference, however, because most, if not all, mathematical functions are implemented using logical circuits.

The TP-900 is, essentially, four TP-600's combined into a single system. Each of the four TP-600's renders an image that is slightly larger than one quarter of the fingerprint image using the process described above. After each of the four images has been rendered by the respective TP-600, the resulting images are combined or merged into a single composite image. The manner in which the four images are combined differs between the two commercial versions of the TP-900. In one version, the TP-900 bisects the overlapping regions, discarding the overlapping data, and butting the image portions together. In the second version, the TP-900 uses a mathematical function to blend the images at their edges and smooth small discontinuities in ridge edges at the boundaries of the quadrants.

3.

DBI sued Identix and Fowler (collectively "Identix") for patent infringement. DBI asserted five claims — claims 1, 3, 16, 17 and 20. Identix moved for partial summary judgment that the TP-600 did not infringe any of the asserted claims. DBI cross-moved for summary judgment. The district court granted Identix's motion and denied DBI's. The court construed the claims to require a data structure in memory capable of storing digital data representing a two-dimensional array. Under this construction, the trial court concluded that the TP-600 did not literally infringe because it did not "store a separate image array in memory for any frame after the first," since it stored only one pixel value at a time. The court also held that prosecution history estoppel precluded a finding of infringement of the TP-600 under the doctrine of equivalents. Accordingly, the court entered summary judgment in favor of Identix.

Identix and DBI then cross-moved for partial summary judgment on the issue of whether the TP-900 infringed. The district court again ruled in favor of Identix and against DBI. The court construed the claims to require that the intermediate arrays used to construct the composite array be representative of an actual image that appeared on the platen. Since the four partial images generated by the TP-900 were themselves composite images, the court concluded that the TP-900 did not satisfy this limitation. DBI unsuccessfully argued that in any case there existed a genuine question of fact as to whether the TP-900's CCDs produced the equivalent of the images characteristic of an actual fingerprint. Further, as with the TP-600, the court concluded that infringement by equivalents of the TP-900 was precluded because of prosecution history estoppel. The court then entered judgment in favor of Identix. DBI now appeals both summary judgment rulings.

B. DISCUSSION

The district court had jurisdiction over this case under 28 U.S.C. § 1338(a) as a case "arising under any Act of Congress relating to patents." We have jurisdiction over final decisions of a district court in such cases. See 28 U.S.C. § 1295(a)(1) (1994).

We review the district court's grant of summary judgment in favor of Identix to determine whether any genuine issues of material fact are in dispute, and whether any errors of law were made. See Quantum Corp. v. Rodime, PLC, 65 F.3d 1577, 1580, 36 USPQ2d 1162, 1165 (Fed. Cir. 1995).

On appeal, DBI argues that the district court erred as a matter of law in its construction of the term "array" in claim 16 by requiring that the data be (1) digital and (2) stored in memory. DBI argues that such an interpretation is contrary to both intrinsic and extrinsic evidence. Identix supports the district court's interpretation of array. Not satisfied to rely exclusively on the district court's grounds, however, Identix also advances another argument on which to affirm. Identix argues that the proper interpretation of "slice data" in claim 16 provides a "separate and independent" basis on which to affirm. We, accordingly, consider each in turn.

Since this phrase only appears in independent claim 16, and not claim 1, it is difficult to see how this court can affirm on this ground alone. Identix also claims DBI waived any argument based on claim 1. We do not agree. While terse, DBI alleged in its opening brief that claim 1 was also infringed. See Brief of Appellant 32, n. 12. Its argument focuses on infringement, however, rather than claim construction. We will consider that issue in Section B.2.a, infra.

1.

Courts have the "power and obligation to construe as a matter of law the meaning of language used in patent claims." Markman v. Westview Instruments, Inc., 52 F.3d 967, 979, 34 USPQ2d 1321, 1329 (Fed. Cir. 1995) (in banc), aff'd, 116 S. Ct. 1384, 38 USPQ2d 1461 (1996). To determine the proper meaning of claims we first consider the so-called intrinsic evidence, i.e., the claims, the written description, and, if in evidence, the prosecution history. See Vitronics Corp. v. Conceptronic, Inc., 90 F.3d 1576, 1582, 39 USPQ2d 1573, 1576 (Fed. Cir. 1996). Even within the intrinsic evidence, however, there is a hierarchy of analytical tools. The actual words of the claim are the controlling focus. See Thermalloy, Inc. v. Aavid Engineering, Inc., 121 F.3d 691, 693, 43 USPQ2d 1846, 1848 (Fed. Cir. 1997) ("Nonetheless, throughout the interpretation process, the focus remains on the meaning of claim language.") The written description is considered, in particular to determine if the patentee acted as his own lexicographer, as our law permits, and ascribed a certain meaning to those claim terms. If not, the ordinary meaning, to one skilled in the art, of the claim language controls. See York Prods., Inc. v. Central Tractor Farm Family Ctr., 99 F.3d 1568, 1572, 40 USPQ2d 1619, 1622 (Fed. Cir. 1996) ("Without an express intent to impart a novel meaning to claim terms, an inventor's claim terms take on their ordinary meaning."). The prosecution history is relevant because it may contain contemporaneous exchanges between the patent applicant and the PTO about what the claims mean. See Vitronics, 90 F.3d at 1582-83, 39 USPQ2d at 1577. If upon examination of this intrinsic evidence the meaning of the claim language is sufficiently clear, resort to `extrinsic' evidence, such as treatises and technical references, as well as expert testimony when appropriate, should not be necessary.

However, if after consideration of the intrinsic evidence there remains doubt as to the exact meaning of the claim terms, consideration of extrinsic evidence may be necessary to determine the proper construction. If a claim falls into this latter category, however, another claim construction canon comes into play. Because the applicant has the burden to "particularly point out and distinctly claim the subject matter which the applicant regards as his invention," 35 U.S.C. § 112, at 2 (1994), if the claim is susceptible to a broader and a narrower meaning, and the narrower one is clearly supported by the intrinsic evidence while the broader one raises questions of enablement under § 112, at 1, we will adopt the narrower of the two. See Athletic Alternatives, Inc. v. Prince Mfg., Inc., 73 F.3d 1573, 1581, 37 USPQ2d 1365, 1372 (Fed. Cir. 1996); Ethicon Endo-Surgery, Inc. v. United States Surgical Corp., 93 F.3d 1572, 1581, 40 USPQ2d 1019, 1026 (Fed. Cir. 1996).

Of course, at some point a lack of clarity may result in a conclusion that the claim is too indefinite to be valid. See 35 U.S.C. § 112, at 2.

After all of these sources are properly considered, a district court must exercise its power and duty to say what the claims mean. We review a district court's claim construction anew and without deference. See Cybor Corp. v. FAS Techs., Inc., 138 F.3d 1448, 1456, 1998 WL 134028, *6, 46 USPQ2d 1169, 1174 (Fed. Cir. 1998) (in banc) ("[W]e review claim construction de novo on appeal including any allegedly fact-based questions relating to claim construction.").

Since the interpretation of claims, as with the interpretation of statutes and contracts, turns on the actual wording of the claim, we consider that first. The two asserted independent claims read:

1. A method of generating data characteristic of a rolled fingerprint image, including:

providing an optical device having a finger receiving surface;

rolling a finger across the finger receiving surface of the optical device and propagating fingerprint images of finger portions in contact with the surface from the device;

imaging the finger receiving surface of the optical device and generating digital data representative of fingerprint images propagated therefrom;

storing arrays of digital data characteristic of adjacent and overlapping fingerprint images of portions of the finger as the finger is rolled across the finger receiving surface of the optical device; and

generating a composite array of digital data characteristic of a rolled fingerprint image as a mathematical function of overlapping image data from a plurality of arrays and characteristic of the overlapping portions of the fingerprint images.

16. A method for generating data characteristic of a rolled fingerprint image, including:

generating arrays of slice data characteristic of adjacent and overlapping two-dimensional slices of the fingerprint image; and

generating a composite array of data characteristic of the rolled fingerprint image as a mathematical function of overlapping slice data from a plurality of overlapping slices.

As with many patent cases, at issue is the meaning of only a few of the words in the claims. In particular, we are asked by appellants to construe "arrays" in claim 16, while appellees request that we interpret "slice data." We consider each.

a.

The word "array" is used repeatedly in the specification. Claim 16 itself uses the word twice: once to refer to the data structures containing the "slice data" and another time to refer to the data structure containing the "data characteristic of the rolled fingerprint image," i.e., the composite array. Thus, whatever interpretation we assign should encompass both uses because the same word appearing in the same claim should be interpreted consistently. See Fonar Corp. v. Johnson Johnson, 821 F.2d 627, 632, 3 USPQ2d 1109, 1113 (Fed. Cir. 1987) (the meaning of a term in a claim must be defined in an manner that is consistent with its appearance in other claims in the same patent).

The fact that the word "array" is used to refer to both the data structures containing the slice data and the composite data supports the district court's claim construction. The pixel values stored in the image "arrays" (e.g., "IAL") are described as "digital values representative of the intensity of optical fingerprint images. . . ." Col. 4, ll. 13-14. Moreover, according to the written description, the "optical image OIL will be propagated from the prism and data representative thereof will be stored in RAM 14 as image array IAL." Thus, the written description teaches that the array is created upon storing the image data in memory.

The composite "array" is also described as containing digital data. This follows naturally from the fact that the data in the composite array is taken directly from the active area of the image array, which is digital. The data in the composite "array" is also stored in memory. See Col. 6, ll. 20-21 ("Composite array CA is stored in RAM 14.") Our interpretation of "slice data," discussed infra, also supports the district court's construction. Accordingly, both the claim language itself and the written description support the district court's claim construction.

DBI points to one isolated passage in the written description to support its claim construction. That passage states that the "[d]igitizer 24 produces two-dimensional arrays of digital pixel values PVn, m/ representative of the intensity of fingerprint images at corresponding discrete pixel locations PLn, m/." It is clear from the entirety of the written description that this is not an accurate statement. The digitizer is an analog-to-digital converter. Analog data comes in and digital data goes out. The digitizer does not organize the data nor index it. Thus, the digitizer does not "produce [an] array" even under DBI's proposed definition of array, which requires the data to be "individually addressable." This isolated passage therefore does not alter our construction, which is based on the entire written description.

In light of the explicit teaching of the `976 patent, it is difficult to understand how the composite data structure could be assembled without storing digital data. The composite image must be stored in a memory separate from the memory used to contain the current image array or otherwise the current image array would overwrite the composite image. Both the current image array and the composite image must also be operated upon in order to integrate the current slice data into the composite image. This requires the data in the composite array to be compared with the image in the active area, which may then be combined according to a mathematical formula. In view of the uncertainties surrounding the implementation of such an invention, we adopt the narrow claim construction that is clearly supported by the written description, and interpret "array" in claim 16 to mean a data structure stored in memory that is representative of a two-dimensional image. See Athletic Alternatives, 73 F.3d at 1581, 37 USPQ2d at 1372.

DBI argues that the plain meaning of the word "array," as evidenced by its dictionary definition, does not require the data contained therein to be digital. To support its position it offers the following definition: "[a]n n-dimensional ordered set of data items identified by a single name and one or more indices, so that each element of the set is individually addressable" (quoting The New IEEE Standard Dictionary of Electrical and Electronic Terms (5th ed. 1993)). The intrinsic evidence clearly indicates that the applicant distinguished between analog "frames" and digital "arrays." Because the intrinsic record is clear we do not give weight to an inconsistent dictionary definition.

b.

Identix argues that properly construed the term "slice data" provides an alternative ground on which to affirm the district court's judgment. See Ethicon, 93 F.3d at 1582, 40 USPQ2d at 1027 (appellate court "must affirm the decision of the district court if it is supported by any ground properly preserved on appeal"); Mark I Mktg. Corp. v. R.R. Donnelley Sons Co., 66 F.3d 285, 289, 36 USPQ2d 1095, 1098 (Fed. Cir. 1995) (appellate court can affirm on alternative basis that is "purely legal and was fully briefed to the district court").

Identix argues that the phrase "slice data" is synonymous with "active area." According to Identix, because the TP-600 does not generate a two-dimensional sub-set of the full image by filtering some of the image data before merging the remaining data into the "composite array," in the words of the patent, the TP-600 does not "generat[e] arrays of slice data" as required by claim 16. Therefore, the TP-600 does not infringe.

DBI, on the other hand, offers the following definition: "A slice is a broader term than active area. A slice is each piece which is used to make up the final rolled fingerprint, and slice data is the data in that slice." Reply Brief of Appellant 9. Furthermore, DBI argues that an active area is a subset of the slice data "that is calculated to approximate the actual slice's data." Id. DBI distinguishes "slice data" from what it calls in its brief "blue sky data," which is the data in the image array that corresponds to that portion of the image propagating surface that is not in contact with any part of the finger. DBI does not argue that the image array itself is an array of slice data.

The written description supports the conclusion that a slice is an active area and the data therein is "slice data." "Slice" is used synonymously with "active area" in two instances in the written description. See col. 6, ll. 37-38 ("Following the copying of the first `slice' or active area"); col. 7, ll. 16-17 ("the current active area or slice being processed"). The written description also speaks of the overlap between the current active area and the composite array in terms of "slice data": "In the preferred embodiments, the composite array is generated as a mathematical function of a comparison and an average of the overlapping slice data characteristic of the overlapping portions of adjacent slices." Col. 2, ll. 15-19 (emphasis added). The only description given is that of merging the overlapping portions of the active areas. Thus, the written description fully supports the conclusion that "slice data" in claim 16 refers to the data in the active area. See also col. 7, ll. 34-37 ("[P]rocessor 12 requires a predetermined minimum overlap such as eight pixel values PVn, m/ between the active areas of the current image array and the composite array.")

In contrast, DBI's proposed claim construction finds no support in the written description. Nowhere is there a description of identifying merely the non-blue-sky data and merging that data into the composite array. The patent only describes merging active areas into the composite array. Moreover, absent some form of processing, the system has no way to distinguish between blue-sky data and non-blue-sky data. The only processing that is described is defining the left, right, top, and bottom boundaries of the active area, not some amorphous combination of pixels in the image array, as DBI's claim construction would require.

DBI finds support for its position in the other claims, which explicitly use the phrase "active area." The argument is that because "active area" was used in some of the claims and not others, the term must have a meaning unique to itself. This argument might have some force if the claims that used "active area" actually depended from claim 16, but none do. As used here, we disagree with DBI's premise. Just because certain words are used in different claims does not mean that those terms cannot have application elsewhere in the patent. See Tandon Corp. v. United States Int'l Trade Comm'n, 831 F.2d 1017, 1024, 4 USPQ2d 1283, 1288 (Fed. Cir. 1987).

The prosecution history supports this reading. Claim 16, then 19, was originally presented in the preliminary amendment. It initially read:

19. A method for generating data characteristic of a rolled fingerprint image, including:

generating arrays of data characteristic of adjacent and two-dimensional slices of the fingerprint images; and

merging the arrays of data into a composite array of data characteristic of the rolled fingerprint image.

That claim was rejected under 35 U.S.C. § 102(b) "as being anticipated by Ruell (German Patent No. 3423886 A1)." (Ruell describes a system much like the present system in that it records a digital representation of a fingerprint image, but it does so in a different way.)

In response to that rejection, the applicants amended the claim as follows (with words added underlined and words deleted in square brackets):

19. A method for generating data characteristic of a rolled fingerprint image, including:

generating arrays of slice data characteristic of adjacent and over-lapping two-dimensional slices of the fingerprint images; and

[merging the arrays of data into] generating a composite array of data characteristic of the rolled fingerprint image as a mathematical function of overlapping slice data from a plurality of overlapping slices.

In the accompanying remarks, applicants' attorney distinguished Ruell, with our emphasis added:

Unlike the applicants' claimed invention, there is no provision for generating the composite array as a function of overlapping image portions of the finger. Similarly, Ruell does not teach a system which identifies active portions of the image which are representative of the fingerprint features as a mathematical function of the stored data, and generates the composite array as a mathematical function of the data representative of the identified active portions.

Appellants ask us to parse through the prosecution history and limit these remarks to those claims in which "active area" explicitly appears. The remarks are not as limited in scope as DBI suggests. The remarks contained in the block quote recited above were made without reference to a particular claim. Instead, the remarks were made with respect to "all of the pending claims [that] stand rejected under 35 U.S.C. § 102(b) or 35 U.S.C. § 103 over the Ruell German Patent 3,432,886." While it is true that the applicants went on to specifically distinguish each claim, or group of claims, including claim 19, from Ruell on more narrow grounds, that does not eliminate global comments made to distinguish the applicants' "claimed invention" from the prior art.

The public has a right to rely on such definitive statements made during prosecution. Notice is an important function of the patent prosecution process, as reflected by the statute itself, see 35 U.S.C. § 112, at 2, and recently confirmed by the Supreme Court, see Warner-Jenkinson Co. v. Hilton Davis Chem. Co., 117 S. Ct. 1040, 41 USPQ2d 1865 (1997). Absent qualifying language in the remarks, arguments made to obtain the allowance of one claim are relevant to interpreting other claims in the same patent. Cf. Southwall Techns., Inc. v. Cardinal IG Co., 54 F.3d 1570, 1579, 34 USPQ2d 1673, 1679 (Fed. Cir. 1995) ("[A]rguments made during prosecution regarding the meaning of a claim term are relevant to the interpretation of that term in every claim of the patent absent a clear indication to the contrary.").

Even were we to agree with DBI that its claim construction was plausible, we would still be compelled under Athletic Alternatives to adopt Identix's claim construction. Identix's narrow construction is clearly supported, if not compelled, by the intrinsic evidence. As in Athletic Alternatives, were we to construe claim 16 in the manner suggested by DBI, we are not sure the resulting claim would be enabled. In fact, this case is even more compelling than Athletic Alternatives because in that case the only question was whether the distance between strings of a tennis racket could be varied in a particular manner, not a particularly difficult technological step. Just the opposite is true in this case. To adopt DBI's claim construction would require the steps of distinguishing between the blue-sky data and the non-blue sky data, storing those data in an array, and combining some, but possibly not all of those data. The manner in which such steps would be performed are not readily apparent and, thus, raise precisely the same question of notice as the court faced in Athletic Alternatives. Accordingly we reject DBI's claim construction on that ground as well.

2.

Having construed the claims, we next consider whether the properly construed claims read on the accused devices. See Markman, 52 F.3d at 976, 34 USPQ2d at 1326. As described above, there are two accused devices in this case — the TP-600 and the TP-900. DBI focuses most of its attention on appeal on the TP-600. Its primary argument for infringement of the TP-900 is that it is comprised of four TP-600s. DBI also argues that the TP-900 infringes because the images generated by each of the four TP-600s overlap, and that the TP-900 combines the overlapping data using a mathematical function. We will consider each product separately.

a. TP-600 1. Literal Infringement.

DBI argues that, when claim 16 is properly interpreted, the TP-600 infringes. In particular, DBI argues that the first step is satisfied by the imaging device, i.e., the CCD. We disagree. First, the data generated by the CCD is not digital data; it is analog. Second the data generated by the CCD is neither "adjacent and overlapping" nor "slice data." The data that the CCD generates represents a complete image, albeit representing only that part of the finger that is currently in contact with the image propagating surface. Thus, the data structure that the CCD generates is not "adjacent" to the subsequent data structure because the two are completely "overlapping." The claim requires both. Accordingly, the TP-600 does not literally infringe.

The CCD also does not produce "slice data." It is undisputed that the TP-600 does not distinguish between image data and blue-sky data before it is merged into the composite array. All of the image data in the TP-600 is merged directly into the composite array. Because the TP-600 does not "generate arrays of slice data," as that phrase has been properly interpreted, it cannot and does not infringe claim 16.

Alternatively, DBI argues that if claim 16 is interpreted to require that the data be digital, the TP-600 nonetheless infringes because its A/D converter converts the analog CCD data to digital. We again disagree. The data produced by the A/D converter is a single value. A single value cannot represent "a two-dimensional image" as the word "array" properly requires. Accordingly, the step of "generating arrays of slice data" does not read on the TP-600 A/D converter.

Independent claim 1 has also been asserted and found to be not infringed. While DBI has adequately preserved this issue for appeal, it has not made much of an effort to urge reversal on that ground. Its main argument appears in a footnote in its opening brief. DBI argues that claim 1 is infringed because "[a]t any one point in time, four pixels are in a storage register in the MIN function." We likewise will deal summarily with the matter.

The district court interpreted claim 1 to require the storing step to create "a data structure capable of representing a two-dimensional image." DBI does not challenge this interpretation. The district court further found that the storage performed by the TP-600 prior to the MIN function does not satisfy this limitation. Specifically, the court found that "at any moment in time the registers [of the TP-600] contain values for only a single pixel and . . . a single value cannot constitute an array, much less be one of adjacent and overlapping fingerprint images."

Even when the evidence is viewed in a light most favorable to DBI, as we are required to do in a motion for summary judgment, we cannot say that the district court erred in granting summary judgment. The data stored by the TP-600 does not represent a "two-dimensional image" that is processed by the TP-600. The most that can be said about the four values temporarily saved by the TP-600 is that they represent a fragment of an array, not an array as that term is properly interpreted in claim 1.

2. Equivalent Infringement.

In order to infringe under the doctrine of equivalents, the TP-600 must satisfy each limitation in the claim either literally or by equivalents. The Supreme Court has recently confirmed that each and every limitation must be satisfied. See Warner-Jenkinson, 117 S. Ct. at 1049, 41 USPQ2d at 1871 ("It is important to ensure that the application of the doctrine, even as to an individual element [limitation], is not allowed such broad play as to effectively eliminate that element in its entirety."). An accused device cannot infringe, as a matter of law, if even a single limitation is not satisfied. See Pennwalt Corp. v. Durand-Wayland, Inc., 833 F.2d 931, 934-35, 4 USPQ2d 1737, 1739-40 (Fed. Cir. 1987) (in banc).

The touchstone for determining whether an element in an accused device is equivalent to a claimed limitation is the substantiality of their differences. In order to infringe under the doctrine of equivalents, the element must differ only insubstantially from the asserted claim limitation. See Dawn Equip. Co. v. Kentucky Farms Inc., 140 F.3d 1009, 1015-16, 46 USPQ2d 1109, 1114-15 (Fed. Cir. 1998); Hilton Davis Chem. Co. v. Warner-Jenkinson Co., 62 F.3d 1512, 1518, 35 USPQ2d 1641, 1645 (Fed. Cir. 1995) (in banc), rev'd on other grounds, 520 U.S. 17, 41 USPQ2d 1865 (1997).

In an appropriate case summary judgment is available as the decisional mechanism for making that determination. The Supreme Court has made it clear that "[w]here the evidence is such that no reasonable jury could determine two elements to be equivalent, district courts are obliged to grant partial or complete summary judgment." Warner-Jenkinson, 117 S. Ct. at 1053 n. 8, 41 USPQ2d at 1875 n. 8 (citing Fed.R.Civ.P. 56 and Celotex Corp. v. Catrett, 477 U.S. 317, 322-23 (1986)).

The TP-600 is too different from the patented invention to support a verdict of infringement under the doctrine of equivalents; no reasonable jury could find otherwise. In the TP-600, each digital data value produced by the A/D converter is temporarily held in a register while that value is processed. These values are not accumulated and then stored in memory, as in the claimed invention. The data value produced by the A/D converter is not filtered to eliminate the blue-sky data from the non-blue-sky data, as in the claimed invention. The prosecution history, as discussed above, confirms the significance of the slice data to the invention. DBI distinguished the Ruell reference, inter alia, on the ground that "Ruell does not teach a system which identifies active portions of the image." The A/D converter generates only a single value; therefore, it does not and can not differentiate between individual values, as in the claimed invention. In light of these significant differences, no reasonably jury could find that the TP-600 infringes either claim 1 or 16 under the doctrine of equivalents. Accordingly, summary judgment of non-infringement was properly granted. See Chiuminatta Concrete Concepts, Inc. v. Cardinal Indus., Inc., ___ F.3d ___, 46 USPQ2d 1752, 1758, 1998 WL 239335, *8 (Fed. Cir. 1998) (concluding that "no reasonable jury" could conclude that the accused device infringed under the doctrine of equivalents); Laitram Corp. v. Morehouse Indus., Inc., ___ F.3d ___, 1998 WL 220092, *7 (Fed. Cir. 1998) (same); Dawn Equip., 140 F.3d at 1017, 46 USPQ2d at 1114-15.

While the district court concluded that prosecution history estoppel precludes a finding of infringement under the doctrine of equivalents, in light of our conclusion that no reasonable jury could find the TP-600 satisfies the limitation of "generating arrays of slice data," we do not reach that issue. For our purposes, it is enough to conclude that the prosecution history reinforces the differences between the claimed invention and the accused device.

a. TP-900 1. Literal Infringement.

DBI makes two arguments in favor of finding infringement by the TP-900. The first is that the TP-900 infringes because it includes four TP-600s, which infringe. Because we affirmed the district court's judgment of non-infringement of the TP-600, this argument must fail.

DBI also argues that the TP-900 infringes claim 16 because each TP-600 generates an array of "slice data" and the TP-900 system "generates a composite array of data characteristic of the rolled fingerprint image as a mathematical function of overlapping slice data." The district court interpreted claim 16 to require that the "arrays of slice data" must be "data characteristic of the portion of the finger in contact with the surface of the platen at a particular time." Because each of the partial images was itself a composite image, the district court found that the partial images did not contain "data characteristic of the portion of the finger in contact with the surface of the platen at a particular time," and thus did not satisfy this limitation. Again, DBI does not challenge the district court's claim construction on this point. It is undisputed that the resulting partial images are themselves composite images made up by merging image data from a plurality of successive images. As such, we cannot say that the district court erred in granting summary judgment, even when the evidence is viewed most favorably to DBI.

2. Equivalent Infringement.

DBI does not argue that the TP-900 infringes under the doctrine of equivalents even if there is no literal infringement, other than its argument that the TP-900 infringes because it includes four TP-600s. Thus, infringement of the TP-900 under the doctrine of equivalents stands or (in this case) falls with the TP-600. Because we conclude that the TP-600 does not infringe, the TP-900 also does not infringe under the doctrine of equivalents.

C. CONCLUSION

For the reasons stated herein, the summary judgments of non-infringement are AFFIRMED.

COSTS

Each party to bear its own costs.