`______________________________________________
`
`BEFORE THE PATENT TRIAL AND APPEAL BOARD
`______________________________________________
`
`Samsung Electronics Co., Ltd., and
`Samsung Electronics America, Inc.,
`Petitioners
`
`v.
`
`Image Processing Technologies, LLC,
`Patent Owner.
`
`
`
`PETITION FOR INTER PARTES REVIEW
`OF U.S. PATENT NO. 6,959,293
`
`
`
`U.S. Patent No. 6,959,293
`Petition for Inter Partes Review
`
`TABLE OF CONTENTS
`
`
`Introduction ..................................................................................................... 1
`I.
`II. Grounds for Standing, Mandatory Notices, and Fee Authorization ............... 1
`III. Summary of Challenges .................................................................................. 3
`IV. Overview of The Relevant Technology and ’293 Patent ............................... 4
`V.
`The Invalidating Prior Art .............................................................................. 8
`A.
`International Patent Publication WO 99/36893 (“Pirim”) ................... 8
`B. U.S. Patent No. 5,546,125 to Tomitaka et al. (“Tomitaka”) .............. 13
`C.
`Robert B. Rogers, “Real-Time Video Filtering With Bit-Slice
`Microprogrammable Processors,” Ph.D. Dissertation, New
`Mexico State University (1978) (“Rogers”) ...................................... 17
`D. Alton L. Gilbert et al., “A Real-Time Video Tracking System,”
`IEEE Transactions on Pattern Analysis and Machine
`Intelligence, Vol. PAMI-2, No. 2, January 1980 (“Gilbert”)............. 24
`VI. Level of Ordinary Skill In The Art ............................................................... 28
`VII. Claim Construction ....................................................................................... 29
`VIII. Specific Explanation Of Grounds For Invalidity.......................................... 30
`A. Ground 1: Claims 1, 18, 19, 22, and 29 are rendered obvious
`under 35 U.S.C. § 103 by the combination of Pirim and
`Tomitaka ............................................................................................. 30
`1.
`Reasons to Combine Pirim and Tomitaka ............................... 30
`2.
`Claim 1 ..................................................................................... 32
`3.
`Claim 18 ................................................................................... 40
`4.
`Claim 19 ................................................................................... 49
`5.
`Claim 22 ................................................................................... 49
`6.
`Claim 29 ................................................................................... 51
`B. Ground 2: Rogers in combination with Gilbert renders obvious
`claims 1, 18, 19, 22, and 29 under 35 U.S.C. § 103. .......................... 53
`1.
`Reasons to Combine Rogers and Gilbert ................................. 53
`2.
`Claim 1 ..................................................................................... 54
`
`
`
`
`
`U.S. Patent No. 6,959,293
`Petition for Inter Partes Review
`
`
`
`Claim 18 ................................................................................... 58
`3.
`Claim 19 ................................................................................... 63
`4.
`Claim 22 ................................................................................... 64
`5.
`Claim 29 ................................................................................... 66
`6.
`C. Ground 3: Claims 1, 18, 19, 22, and 29 are rendered obvious
`under 35 U.S.C. § 103 by the combination of Tomitaka and
`Rogers ................................................................................................. 68
`1.
`Reasons to Combine Tomitaka and Rogers ............................. 68
`2.
`Claim 1 ..................................................................................... 71
`3.
`Claim 18 ................................................................................... 76
`4.
`Claim 19 ................................................................................... 79
`5.
`Claim 22 ................................................................................... 79
`6.
`Claim 29 ................................................................................... 82
`IX. Grounds 1, 2, and 3 are not Cumulative ....................................................... 83
`X.
`Conclusion .................................................................................................... 84
`Certification of Word Count ................................................................................... 84
`
`ii
`
`
`
`
`
`
`
`U.S. Patent No. 6,959,293
`Petition for Inter Partes Review
`
`LIST OF EXHIBITS
`
`iii
`
`
`
`1006
`
`1009
`1010
`
`1011
`
`1012
`
`
`
`
`Exhibit No. Description
`1001
`U.S. Patent No. 6,959,293
`1002
`Declaration of Dr. John C. Hart
`1003
`Curriculum Vitae of Dr. John C. Hart
`1004
`Prosecution File History of U.S. Patent No. 6,959,293
`1005
`WO 99/36893, Patrick Pirim and Thomas Binford, “Method and
`Apparatus for Detection of Drowsiness,” published July 22, 1999
`Robert Rogers, “Real-Time Video Filtering with Bit-Slide
`Microprogrammable Processors,” Ph.D. Dissertation, New Mexico
`State University (December 1978)
`U.S. Patent No. 5,546,125 to Tomitaka, et al., issued August 1996
`Alton L. Gilbert et al., “A Real-Time Video Tracking System,”
`IEEE Transactions on Pattern Analysis and Machine Intelligence,
`Vol. PAMI-2, No. 2, January 1980
`Declaration of Susan E. Beck (authenticating Ex. 1006)
`D. Trier, A. K. Jain and T. Taxt, “Feature Extraction Methods for
`Character Recognition-A Survey”, Pattern Recognition, vol. 29, no.
`4, 1996, pp. 641–662.
`M. H. Glauberman, “Character recognition for business machines,”
`Electronics, vol. 29, pp. 132-136, Feb. 1956
`Declaration of Gerard P. Grenier (authenticating Ex. 1008)
`
`1007
`1008
`
`
`
`
`I.
`
`U.S. Patent No. 6,959,293
`Petition for Inter Partes Review
`
`INTRODUCTION
`Pursuant to 37 C.F.R. § 42.100, et seq., Samsung Electronics Co., Ltd.
`
`(“Petitioner” or “Samsung”) hereby petitions the United States Patent and
`
`Trademark Office (the “Office”) to institute an inter partes review of claims 1, 18,
`
`19, 22, and 29 of U.S. Patent No. 6,959,293 (“the ’293 Patent”). The ’293 Patent,
`
`attached as Ex. 1001, is assigned to Image Processing Technologies, LLC (“Patent
`
`Owner”). The ’293 Patent generally relates to a system and method of analyzing
`
`an aural or visual image or event by using histograms. See, e.g., Ex. 1001 at
`
`claims 1, 18, 22, 29. As set forth below, claims 1, 18, 19, 22, and 29 of the ’293
`
`Patent are invalid as obvious over the prior art. This petition presents non-
`
`cumulative grounds of invalidity based on combinations of prior art that were not
`
`before the Office during prosecution. These grounds are each reasonably likely to
`
`prevail, and this petition, accordingly, should be granted on all grounds.
`
`II. GROUNDS FOR STANDING, MANDATORY NOTICES, AND FEE
`AUTHORIZATION
`Grounds for Standing: Petitioner certifies that the ’293 patent is available
`
`for inter partes review and that Petitioner is not barred or estopped from requesting
`
`an inter partes review challenging the patent claims on the grounds identified in
`
`this petition.
`
`Real Party-In-Interest: Samsung Electronics Co., Ltd.; and Samsung
`
`1
`
`
`
`
`Electronics America, Inc.
`
`U.S. Patent No. 6,959,293
`Petition for Inter Partes Review
`
`Notice of Related Matters: Patent Owner has asserted the ’293 patent
`
`against Petitioner in Image Processing Techs., LLC v. Samsung Electronics Co.,
`
`Ltd., et al., Case No. 2:16-cv-00505-JRG (E.D. Tex., filed May 13, 2016).
`
`Patent Owner has also asserted infringement of U.S. Patents Nos. 8,989,445;
`
`8,893,134; 8,805,001; and 7,650,015 in this same litigation and Petitioner is
`
`concurrently filing IPR petitions requesting review of each of these patents.
`
`Lead and Back-Up Counsel:
`
`• Lead Counsel: John Kappos (Reg. No. 37,861), O’Melveny & Myers
`
`LLP, 610 Newport Center Drive, 17th Floor, Newport Beach, California
`
`92660. (Telephone: 949-823-6900; Fax: 949-823-6994; Email:
`
`jkappos@omm.com.)
`
`• Backup Counsel: Nick Whilt (Reg. No. 72,081), Brian M. Cook (Reg.
`
`No. 59,356), O’Melveny & Myers LLP, 400 S. Hope Street, Los
`
`Angeles, CA 90071. (Telephone: 213-430-6000; Fax: 213-430-6407;
`
`Email: nwhilt@omm.com, bcook@omm.com.
`
`Service Information: Samsung consents to electronic service by email to
`
`IPTSAMSUNGOMM@omm.com. Please address all postal and hand-delivery
`
`correspondence to lead counsel at O’Melveny & Myers LLP, 610 Newport Center
`
`Drive, 17th Floor, Newport Beach, California 92660, with courtesy copies to the
`
`2
`
`
`
`
`email address identified above.
`
`U.S. Patent No. 6,959,293
`Petition for Inter Partes Review
`
`Fee Authorization: The Office is authorized to charge an amount in the
`
`sum of $23,000 to Deposit Account No. 50-0639 for the fee set forth in 37 CFR §
`
`42.15(a), and any additional fees that might be due in connection with this Petition.
`
`III. SUMMARY OF CHALLENGES
`Petitioner respectfully requests cancellation of claims 1, 18, 19, 22, and 29
`
`on the following grounds:
`
`• Ground 1: International Patent Publication WO 99/36893 (“Pirim”) in
`
`combination with U.S. Patent No. 5,546,125 (“Tomitaka”) renders obvious
`
`claims 1, 18, 19, 22, and 29 under 35 U.S.C. § 103;
`
`• Ground 2: Robert B. Rogers, “Real-Time Video Filtering With Bit-Slice
`
`Microprogrammable Processors,” Ph.D. Dissertation, New Mexico State
`
`University (1978) (“Rogers”) in combination with Alton L. Gilbert et al., “A
`
`Real-Time Video Tracking System,” IEEE Transactions on Pattern Analysis
`
`and Machine Intelligence, Vol. PAMI-2, No. 2, January 1980 (“Gilbert”)
`
`renders obvious claims 1, 18, 19, 22, and 29 under 35 U.S.C. § 103; and
`
`• Ground 3: Tomitaka in combination with Rogers renders obvious claims 1,
`
`18, 19, 22, and 29 under 35 U.S.C. § 103.
`
`See Ex. 1002, Hart Decl. ¶¶ 36-37.
`
`3
`
`
`
`
`IV. OVERVIEW OF THE RELEVANT TECHNOLOGY AND ’293
`PATENT
`
`U.S. Patent No. 6,959,293
`Petition for Inter Partes Review
`
`The ’293 patent was filed on February 23, 2001, names Patrick Pirim as the
`
`inventor, and claims priority to a foreign application filed February 24, 2000. Ex.
`
`1001 at 1, 50. It is directed to using histograms for image processing, which was
`
`well known for decades before its priority date. Ex. 1002, Hart Decl. ¶¶19-26. It
`
`claims a device and method for processing a scene by acquiring one or more
`
`histograms of parameters associated with a digitized picture element or “pixel.”
`
`See, e.g., Ex. 1001, ’293 Patent, at claims 1, 18, 22, 29. For example, an input
`
`video signal S(t) comprises a succession of frames, each made up of pixels. Id. at
`
`7:55-63. “This signal S(t) carries a value aij of the parameter A for each pixel (i,
`
`j).” Id. at 7:59-60. Parameter A refers to a property of a pixel, such as its speed,
`
`shape, color, etc. See id. at 1:18-20, 29-31. The values of A for a given frame are
`
`analyzed using a histogram processor, such as depicted in Figure 3, annotated
`
`below:
`
`4
`
`
`
`
`
`U.S. Patent No. 6,959,293
`Petition for Inter Partes Review
`
`
`
`Digital DATA(A), corresponding to parameter A, flows through input
`
`multiplexer 105 (shaded green) to the address input of histogram memory 100
`
`(shaded red). If each DATA(A) were an 8-bit value representing pixel brightness
`
`(ranging from 0 to 255) for a pixel in the frame, the histogram memory would
`
`increment the value stored at the address representing the brightness value for that
`
`pixel. In other words, once the frame is processed, the histogram memory would
`
`contain a value at each of 256 memory addresses representing the number of pixels
`
`having the brightness value corresponding to that address. See id. at 8:45-64. Ex.
`
`5
`
`
`
`
`1002, Hart Decl. ¶¶27-29.
`
`U.S. Patent No. 6,959,293
`Petition for Inter Partes Review
`
`Figure 3 also depicts a “classifier unit” 101 (shaded blue) that takes a
`
`DATA(A) value as an input and evaluates whether it meets a particular condition,
`
`for example, brightness equal to 203. Ex. 1001at 9:31-34. Other embodiments
`
`include more general classifiers that evaluate whether a data value falls within a
`
`certain range or exceeds a certain threshold. See id. at Figs 12, 13a (classifier 119
`
`evaluates whether data P is greater than condition Q). The output of the classifier
`
`indicating whether or not the condition is met is sent to coincidence bus 111
`
`(shaded yellow). Id. at 9:36-42. Output signals from multiple classifiers
`
`associated with other histogram units (B, C, D, E . . .) may also be present on
`
`coincidence bus 111 and are sent to coincidence unit 102 (shaded purple). Id. at
`
`10:34-40. Ex. 1002, Hart Decl. ¶¶30-31.
`
`The coincidence unit 102 (shaded purple) includes logic that determines
`
`whether a pixel will be added to the histogram memory 100 (shaded red) based on
`
`selected classification conditions. Ex. 1001 at Fig. 3 (validation signal).
`
`Validation signal logic might enable the histogram memory when the brightness
`
`parameter for that pixel is greater than 50, or it might enable the histogram
`
`memory only for those pixels with both brightness greater than 100 and color equal
`
`to red. See id. at 9:36-50. Ex. 1002, Hart Decl. ¶¶30-31.
`
`The classification condition described above is a fixed value. However, the
`
`6
`
`
`
`
`classification condition may also be set automatically based on statistics about the
`
`U.S. Patent No. 6,959,293
`Petition for Inter Partes Review
`
`scene. See id. at 11:9-29. For example, in Figure 13a, annotated below, the
`
`classifier 119 (shaded blue) evaluates whether data P is greater than condition Q.
`
`
`
`Q might be derived from statistics such as RMAX (shaded red), the number
`
`of counts in the highest bin, or NBPTS (shaded orange), the number of pixels in
`
`the histogram, for example. Id. at 10:7-31.
`
`Generally, the classifier may be achieved according to numerous
`
`embodiments, the essential being that it allows to place the parameter
`
`DATA(A) with respect to values or limits statistically determined
`
`over a set of former data DATA(A).
`
`Id. at 13:32-36. For example, the processor might determine the maximum
`
`brightness of the pixels in a frame and set a classifier condition for subsequent
`
`7
`
`
`
`
`frames based on that statistic, such as implementing a classifier that selects pixels
`
`U.S. Patent No. 6,959,293
`Petition for Inter Partes Review
`
`having a brightness less than 80% of the maximum brightness. This classifier, in
`
`conjunction with the validation logic, would then ensure that only those pixels
`
`satisfying this condition are included in the histogram. Ex. 1002, Hart Decl. ¶¶31-
`
`35.
`
`V. THE INVALIDATING PRIOR ART
`A.
`International Patent Publication WO 99/36893 (“Pirim”)
`Pirim, Ex. 1005, was published on July 22, 1999, based on an international
`
`application filed January 15, 1999, designating the U.S. and qualifies as prior art
`
`under at least pre-AIA 35 U.S.C. §§ 102(a), 102(b), and § 119 (“but no patent shall
`
`be granted . . . for an invention which had been . . . described in a printed
`
`publication in any country more than one year before the date of the actual filing of
`
`the application in this country.”). Pirim names one inventor in common with the
`
`challenged ’293 patent but has a different inventive entity. Ex. 1005 at 1. Pirim is
`
`of record in the prosecution history but was never discussed or used in a rejection
`
`by the examiner. Further, the obviousness combination presented here was never
`
`considered during prosecution. Ex. 1004, Prosecution History at 110, 201-231;
`
`237-242.
`
`Pirim discloses a system for detecting whether a driver is falling asleep by
`
`acquiring pictures of the driver and forming histograms to analyze opening and
`
`8
`
`
`
`
`closing of the driver’s eyes. Ex. 1005, Pirim, at 5. Pirim’s image processing
`
`U.S. Patent No. 6,959,293
`Petition for Inter Partes Review
`
`system “receives a digital video signal S originating from a video camera or other
`
`imaging device 13 which monitors a scene 13a.” Id. at 12. “Signal S(PI)
`
`represents signal S composed of pixels PI.” Id. at 13. Each video frame comprises
`
`horizontal scanned lines, each including “a succession of pixels or image points PI,
`
`e.g., a1.1, a1.2, and a1.3 for line l1.1.” Id.
`
`With reference to Figure 14, annotated below, Pirim discloses a histogram
`
`unit having a memory 100 (shaded red). Data(V), representing pixel parameter V,
`
`proceeds through input multiplexer 104 (shaded green) to the address input of
`
`memory 100. Id. at 29. Just as in the ’293 patent, a value stored at the address
`
`corresponding to the value of the input data parameter is incremented to
`
`accumulate a histogram of the parameter. Id.
`
`Pirim further discloses a “classifier 25b” (shaded blue) that receives the
`
`data(V) value and compares it to a “register 106 that enables the classification
`
`criteria to be set by the user, or by a separate computer program.” Id. at 29-30.
`
`9
`
`
`
`
`
`U.S. Patent No. 6,959,293
`Petition for Inter Partes Review
`
`
`
`The output of classifier 25b proceeds to a bus 23 (shaded yellow), which
`
`also carries the output of other classifiers in the system. Id. at 31. These signals
`
`proceed to validation unit 31 (shaded purple). “Each validation unit generates a
`
`validation signal which is communicated to its associated histogram formation
`
`block 24-29. The validation signal determines, for each incoming pixel, whether
`
`the histogram formation block will utilize that pixel in forming it histogram.” Id.
`
`10
`
`
`
`
`at 30. Thus, the operation of the system is summarized as follows:
`
`U.S. Patent No. 6,959,293
`Petition for Inter Partes Review
`
`Thus, using the classifiers in combination with validation units 30-35,
`
`the system may select for processing only data points in any selected
`
`classes within any selected domains. For example, the system may be
`
`used to detect only data points having speed 2, direction 4, and
`
`luminance 125 by setting each of the following registers to “1”: the
`
`registers in the validation units for speed, direction, and luminance,
`
`register 2 in the speed classifier, register 4 in the direction classifier,
`
`and register 125 in the luminance classifier. In order to form those
`
`pixels into a block, the registers in the validation units for the x and y
`
`directions would be set to “1” as well.
`
`Id. at 31.
`
`Pirim also discloses that statistical characteristics of the histogram are
`
`calculated, including “the minimum (MIN) of the histogram, the maximum (MAX)
`
`of the histogram, the number of points (NBPTS) in the histogram, the position
`
`(POSRMAX) of the maximum of the histogram.” Id. at 32. Such statistics may be
`
`used to automatically set limits of the classifiers:
`
`Fig. 13 diagrammatically represents the envelopes of histograms 38
`
`and 39, respectively in x and y coordinates, for velocity data. In this
`
`example, XM and YM represent the x and y coordinates of the maxima of
`
`11
`
`
`
`
`
`U.S. Patent No. 6,959,293
`Petition for Inter Partes Review
`
`the two histograms 38 and 39, whereas la and lb for the x axis and lc
`
`and ld for the y axis represent the limits of the range of significant or
`
`interesting speeds, la and lc being the longer [sic] limits and lb and ld
`
`being the upper limited [sic] of the significant portions of the
`
`histograms. Limits la, lb, lc, and ld may be set by the user or by an
`
`application program using the system, may be set as a ratio of the
`
`maximum of the histogram, e.g., XM/2, or may be set as otherwise
`
`desired for the particular application.
`
`Id. at 36-37 (emphasis added). In other words, among the ways the classification
`
`criterion can be set, it can be set to a statistic derived from the histogram, such as
`
`half of the maximum value (of the velocity data in this example).
`
`Pirim also discloses classifiers based on X and Y position of pixels that
`
`restrict histogram processing to only a particular rectangular region:
`
`In order to process pixels only within a user-defined area, the x-
`
`direction histogram formation unit 28 may be programmed to process
`
`pixels only in a class of pixels defined by boundaries, i.e., XMIN and
`
`XMAX. This is accomplished by setting the XMIN and XMAX
`
`values in a user-programmable memory in x-direction histogram
`
`formation unit 28 or in linear combination units 30-35. Any pixels
`
`outside of this class will not be processed. Similarly, y-direction
`
`12
`
`
`
`
`
`U.S. Patent No. 6,959,293
`Petition for Inter Partes Review
`
`histogram formation unit 29 may be used to process pixels only in a
`
`class of pixels defined by boundaries YMIN and YMAX.
`
`Id. at 35. These X and Y MIN and MAX classification criteria may also be
`
`changed automatically by the system using statistics derived from the histograms:
`
`Because the moving object may leave the bounded area the system
`
`preferably includes an anticipation function which enables XMIN,
`
`XMAX, YMIN, and YMAX to be automatically modified by the
`
`system to compensate for the direction of the target. This is
`
`accomplished by determining values for O-MVT, corresponding to
`
`the orientation (direction) of movement of the target within the
`
`bounded area using
`
`the direction histogram, and
`
`I-MVT,
`
`corresponding to the intensity (velocity) of movement. Using these
`
`parameters, controller 42 may modify the values of XMIN, XMAX,
`
`YMIN, and YMAX on a frame-by-frame basis to ensure that the
`
`target remains in the bounded box being searched.
`
`Id. at 39-40. Thus, Pirim discloses automatic updating of classification criteria
`
`based on statistical data derived from the histograms. See Ex. 1002, Hart Decl.
`
`¶¶42-47.
`
`B. U.S. Patent No. 5,546,125 to Tomitaka et al. (“Tomitaka”)
`Tomitaka, Ex. 1007, was filed July 6, 1994 and issued August 13, 1996. Ex.
`
`13
`
`
`
`
`1007 at 1. Tomitaka qualifies as prior art under at least 35 U.S.C. § 102(b).
`
`U.S. Patent No. 6,959,293
`Petition for Inter Partes Review
`
`Tomitaka was not before the PTO during prosecution of the challenged patent. Ex.
`
`1001, ’293 patent at face; Ex. 1004, Prosecution History at 110, 201-231; 237-242.
`
`Tomitaka discloses a “video signal follow-up processing system for
`
`adaptively tracking to the moving of a subject.” Ex. 1007, Tomitaka, at Abstract.
`
`As illustrated in Figure 1, annotated below, a color video signal from optical
`
`system 1 is digitized in A/D 6, and the color of each pixel is separated into
`
`brightness (Y or L) and chroma (C) signals. Id. at 4:7-16. The chroma color signal
`
`is further demodulated into individual color difference signals R-Y and B-Y to
`
`form three color values: Y, R-Y, and B-Y. Id. at 4:17-32. The color data Y, R-Y,
`
`and B-Y is converted into the HLS color coordinate system and written to image
`
`memory 15 (shaded orange) for processing. Id. at 4:39-50. Brightness and hue are
`
`then processed by two histogram units 19 and 20 (shaded blue and red). Id. at 6:1-
`
`6.
`
`14
`
`
`
`
`
`U.S. Patent No. 6,959,293
`Petition for Inter Partes Review
`
`
`
`Hue histogram signal S13 and brightness histogram signal S14 are sent to
`
`follow-up signal processor 16 (shaded purple). Id.at 5:42-65. The follow-up
`
`signal processor 16 forms “feature patterns” from the hue and brightness
`
`histograms that are compared with reference measurements to track an object:
`
`“they are compared with an image portion in a reference measurement frame so
`
`that the panning and tilting of the lens block 1 is adaptively controlled to always
`
`move the position of a detected measurement frame having an image with the
`
`highest similarity to the signal of the reference measurement frame.” Id. at 6:64-
`
`7:2.
`
`However, “when the values of components of the hue signal HUE and the
`
`15
`
`
`
`
`components of the brightness signal Y are close to the threshold values
`
`U.S. Patent No. 6,959,293
`Petition for Inter Partes Review
`
`corresponding to each sort value, the sort values to be included become uncertain
`
`depending on presence or absence of noise.” Id. at 6:14-19. To address this issue,
`
`certain pixels are prevented from being included in the Hue histogram by logic
`
`involving gate 18 (shaded green), comparator 25 (shaded yellow), and noise
`
`threshold signal S15 from the follow-up signal processor 16 (shaded purple). As
`
`seen in Figure 1, a “noise determination signal S15” is determined by the follow-
`
`up signal processor 16 and sent to comparator circuit 25 (shaded yellow), where it
`
`is compared to the saturation value. Id. at 6:40-49.
`
`When the hue signal HUE detected at the saturation/hue detection
`
`circuit 14 is close to the L axis (shown in FIG. 2), there is a possibility
`
`that the hue signal HUE may not have meaning as information since it
`
`is buried in noise because of having low saturation. Such a
`
`meaningless hue signal HUE is removed in the gate circuit 18.
`
`Id. at 6:50-55. Thus, pixels that fail the classification condition set up in
`
`comparator 25 are prevented by logic in gate 18 from being included in the HUE
`
`histogram formed by histogram unit 19. Furthermore, the classification criterion
`
`for rejecting such a pixel, represented by signal S15 from the follow-up signal
`
`processor 16, is based on histogram data inputs S13 and S14. See Ex. 1002, Hart
`
`Decl. ¶¶48-50.
`
`16
`
`
`
`
`
`U.S. Patent No. 6,959,293
`Petition for Inter Partes Review
`
`C. Robert B. Rogers, “Real-Time Video Filtering With Bit-Slice
`Microprogrammable Processors,” Ph.D. Dissertation, New
`Mexico State University (1978) (“Rogers”)
`
`The Rogers dissertation, Ex. 1006, was catalogued with the New Mexico
`
`State University library in 1978. Ex. 1009, Beck Decl. ¶¶ 3-10. Rogers qualifies
`
`as prior art under at least 35 U.S.C. § 102(b). Rogers was not before the PTO
`
`during prosecution of the challenged patent. Ex. 1001, ’293 patent at face; Ex.
`
`1004, Prosecution History at 110, 201-231; 237-242.
`
`Rogers describes a system for tracking a missile or similar object by
`
`digitizing video images and analyzing the pixel intensity (brightness) using
`
`histograms:
`
`Real-time video filtering is concerned with the separation of a target
`
`image from the background scene at standard video data rates. The
`
`scene in the field-of-view (FOV) of the television camera is digitized
`
`to form an n-by-m matrix representation of the picture P as
`
`P = Pij, i = 1, 2, . . . n, j = 1, 2, . . . m,
`
`where Pij represents the pixel intensity at point (i,j).
`
`Ex. 1006, Rogers at 25. Within the camera’s field-of-view, the system defines a
`
`tracking windows comprising three regions—the target region (TR), the plume
`
`region(PR), and the background region (BR), as illustrated in Figure 1 of Rogers
`
`(annotated below):
`
`17
`
`
`
`
`
`U.S. Patent No. 6,959,293
`Petition for Inter Partes Review
`
`
`
`The target region (TR) of the tracking window contains the entire target plus part
`
`of the target’s plume (bright exhaust feature), plus a region of background that is
`
`neither target nor plume. Id. at 26. The plume region (PR) samples the pixel
`
`intensities in the target’s plume, and the background region (BR) samples pixel
`
`intensities that are in the background of the scene. Id.
`
`18
`
`
`
`Rogers discloses that two independent windows that can be used either to
`
`U.S. Patent No. 6,959,293
`Petition for Inter Partes Review
`
`
`
`track two objects, to redundantly track an object and its plume, or to provide a
`
`redundant tracking window (possibly with a different size) for a single target. Id.
`
`at 28-29. This is illustrated in Figure 2, reproduced below:
`
`
`
`For every video field (half of an interlaced frame) acquired, histograms of
`
`the pixel intensity for the pixels falling within each region are acquired: “During
`
`each field, the feature histograms are accumulated for the three regions.” Id. at 29.
`
`Thus, for two tracking windows, six histograms of pixel intensity are formed—TR,
`
`BR, and PR histograms for each of the two tracking windows. The regions are
`
`19
`
`
`
`
`configured such that “1. the BR contains only background points (state of nature
`
`U.S. Patent No. 6,959,293
`Petition for Inter Partes Review
`
`B), 2. the PR contains background and plume points (states of nature B and P), and
`
`3. the TR contains background, plume, and target points (states of nature B, P, and
`
`T)” Id. at 41.
`
`The pixel intensity data used to build the histograms is subjected to a
`
`“threshold classifier” that classifies each pixel in the TR region as target, plume, or
`
`background by comparing it to classification thresholds. Id. at 34-35. For
`
`example:
`
`If XT is the set of pixel values that are to be classified as target pixels,
`
`then an arbitrary pixel value x is classified as
`
`x ∈ XT iff hBR(x) < tBR and hPR(x) < tPR.
`
`Similarly, if XP is the set of pixels which are to be classified as plume
`
`pixels, then
`
`x ∈ XP iff hBR(x) < tBR and hPR(x) > tPR.
`
`The third possibility of classification is the background set XB, which
`
`is chosen if neither one of the above conditions apply:
`
`x ∈ XB iff hBR(x) >= tBR.
`
`Id. at 35. Here, hR(x) is the normalized feature histogram of region R, where R is
`
`the background, plume, or target region and tBR and tPR are classification
`
`thresholds. Id. at 30, 32. The classification thresholds tBR and tPR are not static but
`
`20
`
`
`
`
`instead are set dynamically depending on the data collected in the feature
`
`U.S. Patent No. 6,959,293
`Petition for Inter Partes Review
`
`histograms:
`
`A convenient feature of the thresholding algorithm is that the
`
`threshold value may be adjusted dynamically on a frame-to-frame
`
`basis using a hueristic rule. One intuitively reasonable rule is that the
`
`number of points in XT should be no more than η percent of the
`
`number of points in the TR region, with typical values of η being ten
`
`to twenty percent. A similar statement can be made concerning the
`
`number of plume points in XP. By adjusting the threshold values
`
`according to the observed values of η, the sensitivity of the algorithm
`
`to noise in the observed scenes (and to overlap between the target,
`
`plume, and background distributions) may be tailored to the
`
`application.
`
`Id. at 37. Furthermore, the classification thresholds are based on probability
`
`densities that are updated based on histogram data from prior video frames. Id. at
`
`33. A weighting parameter, ω, can be set between 0 and 1 to specify how heavily
`
`prior histogram data should be weighted. For example:
`
`Letting h(i|j) represent the learned estimate of any probability density
`
`function for the ith field using the sampled density functions for all
`
`21
`
`
`
`previous fields up to the jth field, a linear estimator and predictor can
`
`U.S. Patent No. 6,959,293
`Petition for Inter Partes Review
`
`
`
`be expressed as
`
`h(i|i) = ω h(i|i-1) + (1-ω) hi
`
`and
`
`h(i+l|i) = 2 h(i|i) - h(i-1|i-1),
`
`where hi denotes the observed probability density function at the ith
`
`frame. The above equations provide a linear recursive method for
`
`compiling learned density functions. The weighting factor can be used
`
`to vary the learning rate. When ω=0, the learning effect is disabled,
`
`and the measured histograms are used by the predictor. As ω
`
`increases toward one, the learning effect increases and the measured
`
`density functions have a reduced effect.
`
`Id. at 33. Thus, the threshold classifiers can be automatically updated based on
`
`statistical analysis of the measured histograms.
`
`The pixels identified as target pixels using the threshold classifier are then
`
`fed to the “region definition logic (RDL) which defines the locations of the
`
`tracking windows in the camera’s field-of-view.” Id. at 49.
`
`The region definition logic (RDL) combines the pixel clock and
`
`horizontal sync signals with information specifying the required sizes
`
`and locations of the two tracking windows to produce level signals
`
`22
`
`
`
`
`
`U.S. Patent No. 6,959,293
`Petition for Inter Partes Review
`
`indicating the location of the current pixel relative to each tracking
`
`window.
`
`The first block of hardware in the RDL consists of a group of latches
`
`which are loaded by the video processor (just prior to the end of VSX)
`
`with the location of the top left corner, and the height, and the width,
`
`of the two windows.”
`
`Id. at 59. The tracking windows, in turn, select the subsets of pixels within the
`
`field of view that are accumulated in each of the six intensity histograms, as can be
`
`seen