`
`____________________
`
`BEFORE THE PATENT TRIAL AND APPEAL BOARD
`
`____________________
`
`VALEO NORTH AMERICA, INC., VALEO S.A., VALEO GMBH,
`VALEO SCHALTER UND SENSOREN GMBH,
`AND CONNAUGHT ELECTRONICS LTD.,
`
`Petitioners,
`
`v.
`
`MAGNA ELECTRONICS, INC.,
`
`Patent Owner.
`____________________
`
`Case IPR2015-014101
`Patent 8,643,724 B2
`____________________
`
`
`
`
`
`SUPPLEMENTAL DECLARATION OF DR. GEORGE WOLBERG
`
`
`1 Case IPR2015-01414 has been consolidated with this proceeding.
`
`
`
`
`
`I, Dr. George Wolberg, do hereby declare and state, that all statements made
`
`herein of my own knowledge are true and that all statements made on information
`
`and belief are believed to be true; and further that these statements were made with
`
`the knowledge that willful false statements and the like so made are punishable by
`
`fine or imprisonment, or both, under Section 1001 of Title 18 of the United States
`
`Code.
`
`
`
`Dated: May ___, 2016
`
`
`
`
`
`
`
`
`2
`
`
`
`
`
`TABLE OF CONTENTS
`
`I.
`
`INTRODUCTION ........................................................................................... 4
`
`II.
`
`LEGAL STANDARDS ................................................................................... 4
`
`A. Obviousness ............................................................................................ 5
`
`B.
`
`C.
`
`Reasons to Combine ............................................................................... 5
`
`Claim Construction................................................................................. 6
`
`III.
`
`SUPPLEMENTAL OPINIONS ...................................................................... 7
`
`A.
`
`Claim Construction................................................................................. 7
`
`1.
`
`2.
`
`“a synthesized image is generated without duplication of
`objects” ......................................................................................... 8
`
`“approximates a view as would be seen by a virtual
`camera at a single location” ....................................................... 13
`
`B.
`
`The Prior Art Renders the Challenged Claims Obvious. ..................... 16
`
`1.
`
`2.
`
`Yamamoto and Mitsubishi teach a synthesized image with
`minimal multiple exposure and that approximates a view
`from a single location. ................................................................ 17
`
`A PHOSITA would have found it obvious to scale images
`as part of an image synthesis process. ....................................... 24
`
`
`
`3
`
`
`
`
`
`
`
`I.
`
`INTRODUCTION
`
`
`
`1.
`
`I submit this Supplemental Declaration to offer my opinions regarding
`
`the patentability and validity of claims 1, 3–12, 14, 15, 17, 19–52, 54–67, 69–79,
`
`and 81–86 of U.S. Patent No. 8,643,724 (“the ’724 Patent”). More specifically, I
`
`provide my opinions regarding Patent Owner’s Response (Paper 14), Dr. Etienne-
`
`Cummings’s Declaration (Exhibit 2004) and other exhibits submitted by Patent
`
`Owner.
`
`2.
`
`I set forth my professional qualifications and experience and attached
`
`my curriculum vitae with my declaration submitted on June 15, 2015. (Ex. 1020-
`
`21).
`
`3.
`
`In forming the opinions I express in this Supplemental Declaration, I
`
`considered the materials cited in Patent Owner’s Response, Dr. Etienne-
`
`Cummings’s Declaration, and any materials I cite in this Supplemental Declaration.
`
`I have further considered all materials that were cited in my original declaration
`
`submitted on June 15, 2015. I also rely on my years of education, research and
`
`experience, as well as my investigation and study of relevant materials.
`
`II. LEGAL STANDARDS
`
`4.
`
`In expressing my opinions and considering the subject matter of the
`
`claims of the ’724 Patent, I am relying upon certain basic legal principles that
`
`counsel has explained to me.
`
`
`
`4
`
`
`
`
`
`A. Obviousness
`
`5.
`
`It is my understanding that a patent claim is unpatentable if the claimed
`
`invention as a whole would have been obvious to a person having ordinary skill in
`
`the art (“PHOSITA”) at the time of the invention, in view of the prior art in the field
`
`and analogous fields. This means that even if all of the elements of the claim are
`
`not described or disclosed in a single prior art reference, the claim can still be
`
`unpatentable. I understand that in order to prove that a claimed invention is
`
`unpatentable for obviousness, it is necessary to (1) identify the differences between
`
`the claim and particular disclosures in the prior art references, singly or in
`
`combination; (2) specifically explain how the prior art references could have been
`
`combined in order to arrive at the subject matter of the claimed invention; and (3)
`
`specifically explain why a PHOSITA would have been motivated to so combine the
`
`prior art references. I understand the claims are unpatentable if it is more likely
`
`than not that the claims are obvious.
`
`B. Reasons to Combine
`
`6.
`
`To determine whether a claim is obvious, I understand that I may
`
`combine multiple references if a PHOSITA would have had apparent reasons to
`
`combine the references at the time of the alleged invention. I have been informed
`
`that reasons to combine references can include: (a) combining prior art elements
`
`according to known methods to yield predictable results; (b) simple substitution of
`
`
`
`5
`
`
`
`
`
`one known element for another to obtain predictable results; (c) use of a known
`
`technique to improve similar devices (methods, or products) in the same way;
`
`(d) applying a known technique to a known device (method, or product) ready for
`
`improvement to yield predictable results; (e) choosing from a finite number of
`
`identified, predictable solutions that are “obvious to try” and having a reasonable
`
`expectation of success; (f) known work in one field of endeavor may prompt
`
`variations of it for use in either the same field or a different one based on design
`
`incentives or other market forces if the variations are predictable to a PHOSITA;
`
`(g) some teaching, suggestion, or motivation in the prior art that would have led a
`
`PHOSITA to modify the prior art reference or to combine prior art reference
`
`teachings to arrive at the claimed invention.
`
`C. Claim Construction
`
`7.
`
`I understand that the Board has invited the parties to propose claim
`
`constructions for the ’724 Patent claims under the standard established by Phillips
`
`v. AWH Corp., 415 F.3d 1303 (Fed. Cir. 2005). I have been informed by counsel
`
`that under the Phillips standard, a claim term should be interpreted based on its
`
`ordinary and customary meaning, which is the meaning the term would have had to
`
`a PHOSITA at the time of the invention. I understand that to determine the proper
`
`meaning of a claim term, one primarily looks to “intrinsic evidence,” which
`
`includes patent claims, specification, and prosecution history. I understand that a
`
`
`
`6
`
`
`
`
`
`claim term should be interpreted based on the context of the entire claim, without
`
`rendering any claims terms or words meaningless, as well as the specification of the
`
`patent. I further understand that claims generally should not be interpreted to
`
`exclude embodiments or examples of the claimed invention that are set forth in the
`
`specification. I have also been informed by counsel that evidence such as expert
`
`and inventor testimony, dictionaries, and treatises may be relevant to interpretation
`
`of the claim language, although this type of evidence, sometimes called “extrinsic
`
`evidence,” cannot be used to contradict the claim language or the specification.
`
`III. SUPPLEMENTAL OPINIONS
`
`A. Claim Construction
`
`8.
`
`I provide my opinions on the construction of the terms “a synthesized
`
`image is generated without duplication of objects” and “approximates a view as
`
`would be seen by a virtual camera at a single location” as they appear in every
`
`independent claim of the ’724 Patent. Based on my review of the claim language,
`
`the specification, and my understanding of a PHOSITA at the time of the alleged
`
`invention, it is my opinion that the term “a synthesized image is generated without
`
`duplication of objects” should be interpreted as “the synthesized image is generated
`
`with minimum multiple exposure of objects appearing in overlap zones.” Further, it
`
`is also my opinion that a PHOSITA would understand the meaning of
`
`“approximates a view as would be seen by a virtual camera at a single location,” as
`
`
`
`7
`
`
`
`
`
`it is used in this patent, in the plain language in which it is written.
`
`1. “a synthesized image is generated without duplication of objects”
`
`9.
`
`In my opinion, the term “a synthesized image is generated without
`
`duplication of objects” should be interpreted as “the synthesized image is generated
`
`with minimum multiple exposure of objects appearing in overlap zones.” This
`
`construction is consistent with the claim language and specification of the ’724
`
`Patent. By contrast, Patent Owner has proposed in its Response that this term
`
`should be interpreted as “the generated synthesized image shows primarily one
`
`representation (with minimal multiple exposure) of each object in the captured
`
`scene despite the claimed locations of image capture device.” (Paper 14, Patent
`
`Owner Resp. at 11.) In my opinion, the claim language and specification of the
`
`’724 Patent do not support Patent Owner’s proposed construction.
`
`10. As an initial matter, Patent Owner seeks to add new requirements to
`
`the claims such that they require the synthesized image to show “primarily one
`
`representation . . . of each object.” In my opinion, Patent Owner’s proposal
`
`unnecessarily adds a layer of subjectivity and ambiguity to the claims. Indeed,
`
`those of ordinary skill in the art may reasonably disagree as to whether each object
`
`in a given synthesized image is shown in “primarily one representation.” Patent
`
`Owner further applies its proposed construction in a way that requires the complete
`
`resolution of parallax. For example, Patent Owner states that “images that exhibit
`
`
`
`8
`
`
`
`
`
`the effects of unresolved parallax will not be a synthesized image generated without
`
`duplication of objects, as claimed.” (Paper 14, Patent Owner Resp. at 22.)
`
`11. Patent Owner, however, fundamentally confuses the term “without
`
`duplication of objects” with resolution of parallax. For example, duplication of
`
`objects occurs in images synthesized from multiple cameras to the extent the fields
`
`of view of the cameras overlap. As I explained in my original declaration, the field
`
`of view of each camera can be envisioned as a cone that emanates from that camera,
`
`and when there are multiple cameras with overlapping fields of view, these cones
`
`overlap. (Ex. 1020, Wolberg Decl. at ¶ 65.) The result is that cameras with
`
`overlapping fields of view may capture the same objects multiple times, resulting in
`
`multiple exposure. Duplication, however, is different from parallax. For example,
`
`when each of the cameras have a common center of projection, there will be little or
`
`no parallax (Ex. 2004, Etienne-Cummings Decl. at ¶ 44), but given their
`
`overlapping fields of view, the cameras may nonetheless capture the same objects
`
`multiple times resulting in duplication. Thus, in my opinion, the claim language
`
`does not support adding new requirements pertaining to addressing parallax.
`
`12. Moreover, the specification ’724 Patent does not mention parallax or
`
`methods of resolving parallax, and provides only one method of reducing
`
`duplication of objects in the synthesized image, namely adjusting the field of view
`
`of the center (rear) camera to reduce the overlap zones between the fields of view of
`
`
`
`9
`
`
`
`
`
`the side cameras and the center camera. (Ex. 1001, ’724 Patent at 7:11-64.) Again,
`
`the field of view of each camera can be envisioned as a cone that emanates from
`
`that camera. To minimize overlap, the ’724 Patent discloses a method of changing
`
`the radius of those cones. (Ex. 1001 ’724 Patent at 7:11-64.) This technique
`
`reduces overlap because it effectively makes the cone narrower. As the cones are
`
`narrowed, the overlap zones become smaller and more distant from the vehicle. But
`
`this technique can never eliminate overlap zones. Moreover, narrowing the field of
`
`view for the center camera also increases the blind spot areas behind the vehicle
`
`that are not captured by any of the cameras.
`
`13. To illustrate, the overlap zones in Figure 4 of the ’724 Patent
`
`(reproduced below) are annotated in red. Objects that appear in this zone are
`
`captured by the side camera and the center (rear) camera, and therefore would
`
`appear twice in the synthesized image. The area shaded in gray represents blind
`
`zones behind the car that none of the cameras will image.
`
`
`
`10
`
`
`
`
`
`(Ex. 1001 ’724 Patent at Fig. 4 (annotated).)
`
`
`
`14. To reduce the likelihood that an object will appear in this overlap zone,
`
`the specification of the ’724 Patent discloses a method to narrow the field of view
`
`of the center (rear) camera. As shown in revised Figure 4 (below), narrowing the
`
`field of view of the center camera would certainly reduce the overlap zones. But
`
`this technique cannot totally eliminate the overlap zones because the cone from the
`
`center camera will inevitably intersect with the cones from the side cameras. Any
`
`objects in the resulting overlap zones would appear twice in the synthesized image.
`
`As annotated in Figure 4 (below), this technique also increases the blind zones
`
`(showing in grey) behind the vehicle that are not captured by any of the cameras.
`
`(Ex. 1001 ’724 Patent at Fig. 4 (annotated).)
`
`
`
`15.
`
`In support of its proposed construction, Patent Owner asserts that
`
`“[t]he ’724 patent further accounts for duplication of objects and generates a
`
`synthesized image without duplication of objects by processing the image from
`
`
`
`11
`
`
`
`
`
`image capture device 16 differently than the images from image capture devices
`
`14.” (Paper 14, Patent Owner Resp. at 11.) The cited portion of the specification
`
`discloses a method for scaling, or compressing, the image captured by the center
`
`(rear) camera. (Ex. 1001 ’724 Patent at 15:19-26.) This process adjusts the size of
`
`the center image so that all three images appear in approximately the same scale.
`
`16.
`
`Image compression, or scaling, however, has nothing to do with
`
`reducing or eliminating duplication in the synthesized image. Rather, this
`
`technique only addresses the scale changes between the center (rear) camera and the
`
`two side cameras. To illustrate using Figure 7 of the ’724 Patent, an object at
`
`position P2, is closer to the center (rear) camera 16 than the side camera 14. Thus,
`
`this object would appear somewhat larger in the image captured by the center
`
`camera 16. To compensate for this difference, the center image is compressed so
`
`that objects in the image become smaller. As I explain below, this is a well-known
`
`technique in image processing, but does nothing to resolve duplications from the
`
`overlapping fields of view and certainly would not resolve parallax.
`
`(Ex. 1001 ’724 Patent at Fig. 7.)
`
`
`
`12
`
`
`
`
`
`
`
`17. Further, the ’724 Patent specification recognizes that adjusting the field
`
`of view for the center camera can only reduce or minimize, but cannot eliminate,
`
`duplication of objects. For example, the specification explains that the claimed
`
`system discuses “reduce[s] redundant or duplicative images” in the image portions.
`
`(Ex. 1001, ’724 Patent at 7:60-64.) Likewise, the specification explains that the
`
`reduction in overlap zones, and hence “the presentation of redundant information to
`
`the driver,” “may be accomplished to a satisfactory extent.” (Ex. 1001, ’724 Patent
`
`at 7:7-14.) Patent Owner’s proposed construction is not consistent with the
`
`specification because it implies eliminating duplication and parallax, whereas the
`
`specification cannot eliminate duplication and does not addresses parallax.
`
`Therefore, in my opinion, “a synthesized image is generated without duplication of
`
`objects” should be interpreted as “a synthesized image is generated with minimum
`
`multiple exposure of objects appearing in overlap zones.”
`
`2.
`
`“approximates a view as would be seen by a virtual camera
`at a single location”
`
`18.
`
`I also understand that Patent Owner proposes the term “approximates a
`
`view as would be seen by a virtual camera at a single location” should be
`
`interpreted as “a view that appears to be from a single virtual camera without
`
`different image sizes and disjointed boundary lines.” (Paper 14, Patent Owner
`
`Resp. at 13.) Patent Owner proposes to add new requirements to the claim that
`
`would require the view to be “without different image sizes and disjointed boundary
`
`
`
`13
`
`
`
`
`
`lines” in order to suggest a relationship between the claim language and resolution
`
`of parallax. In my opinion, however, a PHOSITA would have understood the
`
`meaning of “approximates a view as would be seen by a virtual camera at a single
`
`location,” as it is used in this patent, in the plain language in which it is written.
`
`That is, the claimed view is from a single virtual camera and does not have to
`
`perfectly or precisely represent the actual view as would be seen from such a virtual
`
`camera. In fact, I find that the specification of the ’724 Patent contradicts Patent
`
`Owner’s proposed interpretation because it discloses examples that would include
`
`disjointed boundary lines or other distortions during typical driving situations.
`
`19. For example, as I explain above and in my original declaration, the
`
`’724 Patent does not at all discuss parallax and does not suggest any requirement
`
`that the claimed system resolve parallax. Rather, claims simply require a
`
`synthesized image that “approximates a view from a single location.” The
`
`specification also explains that claimed view need not be perfect, and explains that
`
`the synthesized image need only “reduce redundant and duplicative images” and
`
`that this “may be accomplished to a satisfactory extent.” (Ex. 1001, ’724 Patent at
`
`7:11-14, 7:60-64.) Thus, I find no support in the ’724 Patent for requiring a view
`
`“without different image sizes and disjointed boundary lines.”
`
`20. To the contrary, the specification of the ’724 Patent provides examples
`
`of camera-based systems that represent the claimed invention and that would result
`
`
`
`14
`
`
`
`
`
`in disjointed boundary lines, different image sizes, and other distortions. In one
`
`example, the disclosed system compresses only the bottom-half of the central
`
`image, whereas the top-half of the center image remains unchanged. (Ex. 1001,
`
`’724 Patent at 15:62-16:4.) The underlying rationale is that objects in the top-half
`
`of the image are far away during most driving situations. Therefore, any difference
`
`in scale between the images captured by the side cameras and the top-half of the
`
`center (rear) camera are expected to be negligible in most driving situations. As the
`
`’724 Patent explains:
`
`In an alternative embodiment, the vertical compression technique may
`
`be applied to only a lower vertical portion of central image portion
`
`48’. In most driving situations, objects imaged by rearward-facing
`
`image capture devices above the horizon are at a long distance from
`
`the vehicle while those below the horizon get progressively closer to
`
`the vehicle in relation to the distance below the horizon in the
`
`displayed image. Therefore, compression of the upper vertical portion
`
`of the central image portion may be eliminated without significant
`
`reduction in performance.
`
`(Ex. 1001, ’724 Patent at 15:62-16:4.)
`
`21. This technique may reduce disjointed boundary lines or different
`
`image sizes in some driving situations, such as when a car is driving down an open
`
`road with few tall structures around. In other common driving situations, however,
`
`the disclosed technique would leave disjointed boundary lines, different image
`
`
`
`15
`
`
`
`
`
`sizes, and other distortions in the synthesized image. For example, the synthesized
`
`image would be particularly distorted if there was a large object directly behind the
`
`vehicle, such as another car or building, which occupies the full field of view of the
`
`center (rear) camera. In this situation, the bottom half of the object would be drawn
`
`at one scale while the top half would be drawn at another scale. The resulting
`
`synthesized image would inevitably include disjointed boundary lines and other
`
`distortions. Nonetheless, the ’724 Patent specification states that such a technique
`
`will generate a single view “without significant reduction in performance.” (Ex.
`
`1001, ’724 Patent at 16:2-4.) Therefore, the specification contemplates that the
`
`synthesized image only needs to approximate the view from a single location and
`
`can include images with some disjointed boundaries or other distortions.
`
`B.
`
`The Prior Art Renders the Challenged Claims Obvious.
`
`22. As I explained in my original declaration, it is my opinion that it would
`
`have been obvious to a PHOSITA to combine the teachings of Yamamoto,
`
`Mitsubishi and Lemelson. I understand the Patent Owner asserts that the
`
`combination of Mitsubishi and Yamamoto would not produce a synthesized image
`
`“without duplication of objects . . . [that] approximates a view as would be seen by
`
`a virtual camera at a single location exterior of the equipped vehicle.” (Paper 14,
`
`Patent Owner Resp. at 28.) Instead, Patent Owner argues that Mitsubishi and
`
`Yamamoto do not meet the claim because they do not disclose how to resolve
`
`
`
`16
`
`
`
`
`
`parallax. (Paper 14, Patent Owner Resp. at 26-27.) But as I explain above, the
`
`claims do not require the elimination of parallax, but only require the synthesized
`
`image to have minimal multiple exposures and approximate a view from a single
`
`virtual camera. In my opinion, Patent Owner’s arguments are incorrect for the
`
`reasons below.
`
`1.
`
`Yamamoto and Mitsubishi teach a synthesized image with
`minimal multiple exposure and that approximates a view
`from a single location.
`
`23.
`
`I understand that Patent Owner criticizes my opinions for “gloss[ing]
`
`over the problem of parallax when asserting that Yamamoto and Mitsubishi disclose
`
`a synthesized image without duplication of objects.” (Paper 14, Patent Owner
`
`Resp. at 26). Patent Owner further asserts that Mitsubishi does not disclose a
`
`synthesized image as claimed in the ’724 Patent and that “Dr. Wolberg admits that
`
`the images must first be properly scaled and aligned, but . . . does not assert that
`
`Mitsubishi discloses how to properly scale and align the images to overcome
`
`Yamamoto’s parallax.” (Paper 14, Patent Owner Resp. at 27.) In my opinion, even
`
`if Yamamoto, Mitsubishi and Lemelson do not expressly address parallax issues,
`
`they still teach a synthesized image with minimal multiple exposure and that
`
`approximates a view from a single location.
`
`24. As I explain above, the claims of the ’724 Patent do not require the
`
`synthesized image to resolve parallax. Nor does the specification disclose any
`
`
`
`17
`
`
`
`
`
`technique for resolving parallax. Instead, the ’724 Patent only teaches how to
`
`minimize duplication of objects in the synthesized image by adjusting the field of
`
`view of the center (rear) camera, as well as aligning the scale of all three images by
`
`compressing the center image. Neither of these techniques resolves parallax. Yet,
`
`according to the specification, the resulting image nonetheless approximates the
`
`view as would be seen by a virtual camera at a single location. Thus, a PHOSITA
`
`would have understood that the claimed synthesized image need not be a perfect
`
`reproduction of the view from a single virtual camera, but rather an approximation
`
`that may contain some artifacts or distortions from parallax.
`
`25.
`
`In my original declaration, I explained that Figure 5 of Yamamoto
`
`depicts an image that mimics the view from the perspective of a camera positioned
`
`exterior to the vehicle. (Ex. 1020, Wolberg Decl. at 47-48.) I understand that
`
`Patent Owner asserts that “Yamamoto does not disclose how the images from these
`
`cameras are combined into a synthesized image that approximates a view as would
`
`be seen by a virtual camera.” (Paper 14, Patent Owner Resp. at 29.) Techniques
`
`for generating the synthesized image shown in Yamamoto, however, were known to
`
`those of ordinary skill in the art and disclosed by Mitsubishi. In particular,
`
`Mitsubishi explains:
`
`The image correcting unit 13 corrects the images in order to join
`
`(synthesize) them by performing a coordinate conversion on data with
`
`the same patterns (for example, PTA and PTA’) retrieved from
`
`
`
`18
`
`
`
`
`
`overlapping portions c, d in P1 image 14 and P2 image 15. . . . The
`
`overlapping regions of P2 image 15 and P3 image 16 are corrected in
`
`the same way to join and synthesize P2 image 15 and P3 image 16
`
`. . . . Because the display areas 1a, 2a, 3a in images P1, P2 and P3
`
`cover the entire field of view from cameras 1, 2 and 3, the desired
`
`portions (denoted by 7’) are included in the series of images that are
`
`displayed on the image display screen 7 by the display output unit 6.
`
`In other words, the area denoted by 7’ in FIG. 3 is displayed on the
`
`image display unit 7.
`
`(Ex. 1005, Mitsubishi at 7, Fig 3.)
`
`
`
`Therefore, Mitsubishi discloses a system that matches the patterns in the
`
`overlapping portions and joins the three separate images. Because the system aligns
`
`the overlap regions, it will minimize any duplication from the overlap zones
`
`between each camera. Mitsubishi further discloses an arithmetic processing unit for
`
`merging the three images, which results in a synthesized image that approximates
`
`the view from a single virtual camera.
`
`
`
`19
`
`
`
`
`
`26.
`
` My opinion is further supported by the declaration of Dr. Etienne-
`
`Cummings. In his declaration, Dr. Etienne-Cummings performs a demonstration in
`
`which he took three photographs at three positions around a vehicle, and merged the
`
`images using various techniques. (Ex. 2004, Etienne-Cummings Decl. at 45-51.)
`
`In my opinion, Dr. Etienne-Cummings’ demonstration shows that a synthesized
`
`image generated using the technique taught by Mitsubishi will approximate the
`
`view from a single virtual camera.
`
`27. According to Dr. Etienne-Cummings, he took pictures from three
`
`locations around a Honda Accord: one at the driver side mirror, another from the
`
`passenger side mirror, and one from the rear of the car. (Ex. 2004, Etienne-
`
`Cummings Decl. at 45-46.) The three images are shown below, respectively.
`
`
`
`
`
`
`
`20
`
`
`
`
`
`
`
`(Ex. 2004, Etienne-Cummings Decl. at 45-46 (top is image from driver side mirror;
`
`middle is image from passenger side mirror; and bottom is image from rear).)
`
`28. Dr. Etienne-Cummings merges the three images in several different
`
`ways and asserts that no arrangement of the three images would result in an
`
`approximation of a view from a single virtual camera. (Ex. 2004, Etienne-
`
`Cummings Decl. at 50-51.) I disagree with Dr. Etienne-Cummings conclusion.
`
`Although none of the merged images are perfect, his demonstration confirms that an
`
`image approximating the view from a virtual camera can be generated using the
`
`technique disclosed in Mitsubishi.
`
`29. For example, in one of his demonstrations, Dr. Etienne-Cummings
`
`creates a synthesized image by scaling the center image and matching the
`
`overlapping portions of each image. (Ex. 2004, Etienne-Cummings Decl. at 45-51.)
`
`The result is reproduced below.
`
`
`
`21
`
`
`
`
`
`
`
`(Ex. 2004, Etienne-Cummings Decl. at 50.)
`
`30. As shown in the following side-by-side comparison, Dr. Etienne-
`
`Cummings’ demonstration mirrors
`
`the
`
`technique disclosed
`
`in Mitsubishi.
`
`Mitsubishi also explains that only “the area denoted by 7’ in FIG. 3 is displayed on
`
`the image display unit 7.” (Ex. 1005, Mitsubishi at 7.) Thus, I have highlighted the
`
`portion of Dr. Etienne-Cummings’ merged image that approximates the area 7’ in
`
`Figure 3 of Mitsubishi.
`
`Fig. 3 of Mitsubishi
`
`Dr. Etienne-Cummings
`
`31. The figure below shows the synthesized composite image from
`
`
`
`
`
`Dr. Etienne-Cummings’ declaration that would be displayed to the driver according
`
`
`
`22
`
`
`
`
`
`to the technique disclosed by Mitsubishi, prior to any processing that may be done
`
`to smooth the images. In this resulting image, there are some minor distortions near
`
`the boundaries between the images. In fact, it appears Dr. Etienne-Cummings has
`
`attempted to complicate the resulting image by using three images taken from a car
`
`parked on a busy street with multiple houses and cars in the immediate background,
`
`even though the specification explains that “[i]n most driving situations, objects
`
`imaged by rearward-facing image capture devices above the horizon are at a long
`
`distance from the vehicle.” (Ex. 1001, ’724 Patent at 15:64-16:2.) Nonetheless, the
`
`resulting image as a whole approximates the view from a single virtual camera and
`
`would provide a useful rearview to the driver of the vehicle. Further, the resulting
`
`image would not have discouraged or otherwise led a PHOSITA to abandon the
`
`three-camera arrangement of Yamamoto. Therefore, Dr. Etienne-Cummings’
`
`declaration confirms my opinion that the technique disclosed by Mitsubishi and
`
`Yamamoto would generate a synthesized image that approximates the view from a
`
`single virtual camera.
`
`32. Further, although Dr. Etienne-Cummings asserts that “[t]he combined
`
`image continues
`
`to have many
`
`instances of
`
`incongruity, distortions and
`
`
`
`
`
`23
`
`
`
`
`
`discontinuities,” he fails to identify anything in the ’724 Patent that resolves such
`
`imperfections. (Ex. 2004, Etienne-Cummings Decl. at 50-51.) Again, the ’724
`
`Patent only discloses compressing the center image, which may help scale the
`
`center image but does not resolve parallax. And the claims specify a synthesized
`
`image that approximates the view from a virtual camera rather than a perfect image.
`
`2.
`
`A PHOSITA would have found it obvious to scale images as
`part of an image synthesis process.
`
`33.
`
`I understand that Patent Owner argues Mitsubishi fails to disclose
`
`scaling the images to produce a synthesized image. (Paper 14, Patent Owner Resp.
`
`at 26-28.) I disagree with Patent Owner’s assertion. In my opinion, a PHOSITA
`
`would have recognized that Mitsubishi teaches a general “image synthesis process”
`
`for combining multiple images was based on the registration, or alignment, of the
`
`images (Ex. 1005, Mitsubishi at 5), and it was well known at the time that scaling is
`
`an important element of image registration (Ex. 1026, Brown at 10). Thus, in my
`
`opinion, a PHOSITA would have found it obvious to scale the images produced by
`
`the three-camera arrangement in Yamamoto to generate a synthesized image as
`
`shown in Figure 5 of Yamamoto.
`
`34. Significantly, Mitsubishi generally
`
`teaches
`
`the use of
`
`image
`
`registration to merge or synthesize the images captured by three different cameras.
`
`For example, Mitsubishi explains that “[t]he image correcting unit 13 corrects the
`
`images in order to join (synthesize) them by performing a coordinate conversion on
`
`
`
`24
`
`
`
`
`
`data with the same patterns . . . retrieved from overlapping portions . . . .” (Ex.
`
`1005, Mitsubishi at 7.) In my opinion, a PHOSITA would have readily understood
`
`that the image correcting unit described in Mitsubishi performs image registration,
`
`i.e., image alignment, to merge the three images.
`
`35. Further, a PHOSITA would have recognized that image registration
`
`techniques for aligning images from several cameras include basic methods such as
`
`translation, scaling and rotation. For example, a highly cited 1992 survey paper on
`
`image registration by Brown et al. describes scaling as one of the most commonly
`
`used Cartesian operations for image matching:
`
`The most commonly used registration transformation is the affine
`
`transformation which is sufficient to match two images of a scene
`
`taken from the same viewing angle but from a different position, i.e.,
`
`the camera can be moved, and it can be rotated around its optical axis.
`
`This affine transformation is composed of the Cartesian operations of
`
`a scaling, a translation, and a rotation.
`
`(Ex. 1026, Brown at 10