`
`
`
`
`
`
`
`
`UNITED STATES PATENT AND TRADEMARK OFFICE
`__________________________________
`
`BEFORE THE PATENT TRIAL AND APPEAL BOARD
`__________________________________
`
`APPLE INC.,
`Petitioner,
`
`v.
`
`RED.COM, LLC,
`Patent Owner.
`__________________________________
`
`Case No. IPR2019-01065
`Patent No. 9,245,314
`__________________________________
`
`DECLARATION OF THOMAS GRAEME NATTRESS IN SUPPORT OF
`PATENT OWNER RED.COM, LLC PRELIMINARY RESPONSE
`
`
`
`
`RED.COM Ex. 2001
`Apple v. RED.COM
`IPR2019-01065
`
`
`
`Apple v. RED.COM
`Declaration of Graeme Nattress re POPR - IPR2019-010164 and IPR2019-01065
`
`
`I, Thomas Graeme Nattress, declare and state as follows:
`
`I. INTRODUCTION
`
`1.
`
`I am a lead camera systems architect at RED.COM (“RED”), the
`
`assignee of U.S. Patent Nos. 9,230,299 (“the ’299 patent”) and 9,245,314 (the ’314
`
`patent). I am also a listed inventor on the ’299 and ’314 patents, and an inventor
`
`on over twenty-five additional issued patents assigned to RED. I am submitting
`
`this declaration in connection with Patent Owner Preliminary Responses to
`
`IPR2019-010164 and IPR2019-01065, filed by Petitioner Apple Inc., relating to
`
`the ’299 and ’314 patents, respectively.
`
`2.
`
`I executed a previous declaration, which was submitted on January 31,
`
`2014, in connection with a reexamination proceeding on U.S. Patent No. 8,174,560
`
`(the “’560 patent”), assigned to RED and for which I am a named inventor (the
`
`“2014 declaration”). A copy of my 2014 declaration can be found at pages 312-
`
`345 in Petitioner’s Ex. 1002 in IPR2019-01064. I understand that Petitioner’s Ex.
`
`1002 in IPR2019-01064 is a copy of the prosecution history of the ’299 patent. A
`
`true and correct copy of my 2014 declaration is also attached hereto as Exhibit
`
`2024 for use and reference in IPR2019-01065, as my 2014 declaration does not
`
`appear in Ex. 1002 (’314 patent prosecution history) in IPR2019-01065.
`
`3.
`
`A detailed summary of my biography and credentials can be found in
`
`my 2014 declaration at paragraphs 1-6.
`
`-1-
`
`
`
`Apple v. RED.COM
`Declaration of Graeme Nattress re POPR - IPR2019-010164 and IPR2019-01065
`
`
`4.
`
`In my 2014 declaration, I testified regarding the inventions claimed in
`
`the ’560 patent. In particular, paragraphs 10-28 in my 2014 declaration describe
`
`the state of the art at the time of the filing of the ’560 patent, and the deficiencies in
`
`technology and conventional thinking that were overcome by the inventions
`
`claimed in the ’560 patent. Those paragraphs relate to the April 11, 2007 time
`
`frame, which is the earliest filing date to which the ’560 patent claims priority.
`
`The ’299 and ’314 patents claim priority to the ’560 patent, and to the same April
`
`11, 2007 provisional application as the ’560 patent. Thus, paragraphs 10-28 in my
`
`2014 declaration refer to the deficiencies in technology and conventional thinking
`
`that were overcome by the inventions in claimed in the ’299 and ’314 patents.
`
`II. REDUCTION TO PRACTICE OF THE
`RED ONE MOTION PICTURE CAMERAS
`
`5.
`
`I first met Mr. Jim Jannard in December of 2005. At that meeting, we
`
`discussed his desire to create the first ever digital motion picture camera that could
`
`record compressed digital motion video at cinema quality levels, including 4K. I
`
`was intrigued by the possibilities, because combining the ease and flexibility post-
`
`production of digital video while maintaining cinema-quality frame rate and
`
`resolution would be a game-changer in the world of movie making. But I also
`
`knew that digital compression was highly disfavored in the movie camera industry
`
`-2-
`
`
`
`Apple v. RED.COM
`Declaration of Graeme Nattress re POPR - IPR2019-010164 and IPR2019-01065
`
`due to its resulting artifacts and lower resolution, which were unacceptable for big-
`
`screen cinema viewing.
`
`6.
`
`To solve this problem, one key area I researched was the use of an
`
`image sensor with a Bayer-pattern filter. This was one of several unconventional
`
`avenues we explored at RED. At the time, such sensors were associated with
`
`lower-quality consumer-grade video cameras and derided as incapable of providing
`
`cinema-quality video due to artifact and resolution issues. In contrast, the industry
`
`consensus held that cinema quality cameras would need to utilize three sensors,
`
`with a prism to split red, green and blue light to each sensor. However, we
`
`believed that the benefits of a Bayer-pattern image sensor could be optimized if the
`
`image data remained in raw, mosaiced format for compression. We believed that
`
`such a data workflow could allow the raw digital files to operate as a digital
`
`negative, and would provide all of the post-production flexibility of being able to
`
`manipulate the original raw data.
`
`Developing REDCODE for the RED ONE Cameras
`
`7.
`
`Immediately following my December 2005 meeting with Mr. Jannard,
`
`I began working on the design of RED’s first commercial digital motion picture
`
`camera that would become known as the RED ONE. This work would last all
`
`throughout 2006 and into 2007 when we commercially launched the RED ONE
`
`-3-
`
`
`
`Apple v. RED.COM
`Declaration of Graeme Nattress re POPR - IPR2019-010164 and IPR2019-01065
`
`video camera. My title on the RED ONE project was Problem Solver, and it
`
`remains my title to this day.
`
`8.
`
`As I explained in paragraphs 13-21 of my 2014 declaration, the
`
`typical flowchart for prior art video image data capture and processing involved
`
`demosaicing video data captured by a Bayer-pattern sensor prior to compressing
`
`that data for storage. Paragraphs 22-28 of my 2014 declaration explain how we
`
`rejected
`
`that conventional
`
`thinking by
`
`implementing known compression
`
`techniques, such as JPEG 2000, in such a way as to operate on non-demosaiced
`
`Bayer-pattern image data, a type of image data these techniques were not designed
`
`to work on.
`
`9.
`
`Accordingly, my work on the RED ONE camera was based on a video
`
`image processing pipeline that did not include a conventional demosaicing step
`
`prior to compression. Rather, the data flow upon which I based my research and
`
`design work for the RED ONE camera generally operated along the following
`
`four-step sequence for processing and compressing raw Bayer-pattern video image
`
`data: First, raw image data was collected by the image sensor, which was a
`
`Mysterium CMOS image sensor chip with a Bayer-pattern pixel filter, i.e., a sensor
`
`chip that detected only one data value for each of the green, red and blue pixel
`
`locations. The raw mosaiced data was outputted by the image sensor at a
`
`resolution of at least 2K and at least 23 frames per second. Second, the raw Bayer-
`
`-4-
`
`
`
`Apple v. RED.COM
`Declaration of Graeme Nattress re POPR - IPR2019-010164 and IPR2019-01065
`
`pattern image data was then sent to a Xilinx processing FPGA chip for pixel
`
`correction and processing of the raw Bayer-pattern image data. Third, the pixel-
`
`corrected and processed raw Bayer-pattern video image data was then sent to
`
`Analog Devices compression chips, which utilized a mathematically lossy wavelet
`
`compression codec known as JPEG 2000, to compress the processed raw Bayer-
`
`pattern video image data. Fourth, the compressed raw Bayer-pattern image video
`
`data was sent to a memory device, by way of a SATA port, for storage as a raw
`
`data file that at one point in development was given a .JIM file extension in
`
`homage to Mr. Jannard and the new file type that RED had created. We referred to
`
`the above programing for the RED ONE cameras, and the resulting raw
`
`compressed data files that it generated, as REDCODE.
`
`10. Much of my research and design work that went into the RED ONE
`
`focused on the second step above, i.e., how to process the raw Bayer-pattern video
`
`image data before compression in way would optimize the decompressed and
`
`demosaiced output. This was a key step in REDCODE because, without pre-
`
`processing, JPEG 2000 compression was otherwise incapable of processing raw
`
`Bayer-pattern image data in a useful way. To solve this problem, I devised and
`
`tested a number of methods for processing the raw Bayer-pattern video image data
`
`prior to compression. I ultimately devised two methods in particular that were
`
`-5-
`
`
`
`Apple v. RED.COM
`Declaration of Graeme Nattress re POPR - IPR2019-010164 and IPR2019-01065
`
`employed on the RED ONE cameras, referred to as GAS (green average
`
`subtraction) and pre-emphasis.
`
`11. The GAS processing technique uses an averaging of green pixel
`
`values to produce low noise, co-sited estimated green values. By way of
`
`background, the highest quality demosaic algorithms work on a large kernel to
`
`provide high quality estimates of edge direction. The edge direction is perturbable
`
`from changes to the data in the input image data to the demosaic algorithm to a
`
`much greater extent than the small local average used in the GAS method. In the
`
`GAS method, green pixels located adjacent or in close proximity to a target pixel
`
`are averaged to calculate an average green value, which is then subtracted from the
`
`value of the target pixel, whether the target pixel is red or blue. For example, a
`
`distinct average green value may be calculated for each target pixel location, then
`
`those distinct average green values are subtracted from the values of the
`
`corresponding distinct red and blue pixels. This subtraction can result in a large
`
`number of zeros, for example, in areas of an image that are black, white or shades
`
`of gray; where the values generated by the image sensor for all three colors are
`
`equal.
`
`12. The pre-emphasis processing technique can be used to help preserve
`
`image detail. This technique helps tune the compression of pixel values through a
`
`pre-emphasis curve that changes the gradient in different sensor code value regions
`
`-6-
`
`
`
`Apple v. RED.COM
`Declaration of Graeme Nattress re POPR - IPR2019-010164 and IPR2019-01065
`
`to either emphasize or de-emphasize those pixel values relative to others in the
`
`image. For example, this technique can be used to spread-apart (increase gradient)
`
`pixel values in some ranges of values and squeeze-together (reduce gradient)
`
`values in one or more other ranges. The spreading-apart of values preferentially
`
`preserves the variations in values, and thus image details, in the ranges where the
`
`values are spread apart so that the corresponding details better survive the
`
`compression process. In one implementation, the pre-emphasis curve is set to
`
`spread apart pixel values corresponding to the darkest image regions, whereas the
`
`relatively brightest image regions were squeezed together. This allowed for
`
`compression that better preserved the details in the darkest areas of the image,
`
`which can otherwise often be washed out. After sufficient testing to find the most
`
`ideal curve, keeping in mind the unfixed rendering intent of the raw image data, I
`
`settled on a simple power law curve with adjustable black offset. This curve
`
`resulted in low complexity and a single tuning variable, among other benefits.
`
`13. The REDCODE programming described above formed the backbone
`
`of the RED ONE cameras, including the two RED ONE cameras known as Boris
`
`and Natasha.
`
`The Boris RED ONE Motion Picture Camera
`
`14. RED’s work refining and debugging the RED ONE continued
`
`throughout 2006 and into 2007. With the goal of debuting the RED ONE at the
`
`-7-
`
`
`
`Apple v. RED.COM
`Declaration of Graeme Nattress re POPR - IPR2019-010164 and IPR2019-01065
`
`upcoming NAB 2007 show taking place in Las Vegas on April 14-19, 2007,
`
`several RED team members tested a RED ONE motion picture camera, nicknamed
`
`Boris, on March 8, 2007 at RED’s Lake Forest, California, headquarters. I was
`
`present for that testing, as was Jim Jannard, Jarred Land, Ted Schilowitz, David
`
`Macintosh and Stuart English, among others. I documented the event by taking
`
`photos with my personal camera. True and correct copies of seven of those photos
`
`are attached hereto as Exhibits 2002-2008.
`
`15. To confirm the dates of the photos that I took of the testing of Boris, I
`
`ran ExifTool on the photo files to collect the metadata for the photos. Attached
`
`hereto as Exhibit 2009 is a true and correct copy of the output of the metadata of
`
`files on my camera, including those relating to the photos those I took of the
`
`testing of Boris at RED headquarters. Exhibits 2002-2008, respectively,
`
`correspond to IMG_4291, IMG_4327, IMG_4317, IMG_4338, IMG_4354,
`
`IMG_4364 and IMG_4365, which all indicate a “Date/Time Original” of March 8,
`
`2007. As an aside, the timestamp associated with the “Date/Time Original”
`
`indication reports time for the Eastern time zone, which corresponds to my home
`
`time zone at the time in Ottawa, Canada.
`
`16. Exhibit 2002, reproduced below, shows Jim Jannard inspecting Boris.
`
`-8-
`
`
`
`Apple v. RED.COM
`Declaration of Graeme Nattress re POPR - IPR2019-010164 and IPR2019-01065
`
`
`17. Exhibit 2003, reproduced below, shows, from left to right, David
`
`Macintosh, Jim Jannard, Ted Schilowitz, Jarred Land and Stuart English standing
`
`next to Boris after it had been set up on a tripod to conduct test footage.
`
`
`
`
`
`-9-
`
`
`
`Apple v. RED.COM
`Declaration of Graeme Nattress re POPR - IPR2019-010164 and IPR2019-01065
`
`
`18. Exhibit 2004, reproduced below, is a close-up shot of Boris on the
`
`tripod. David Macintosh can be seen in the background. This photo also shows a
`
`wire running from Boris to a live monitor (out of frame) on which we viewed the
`
`live feed from Boris. Also visible in this photo is the internal framework for
`
`Boris’s hardware circuitry, as Boris’s cover is removed. This photo also shows
`
`that a section of Boris between the body and lens had a black powder coating,
`
`while the remainder of the exterior was milled aluminum.
`
`
`
`
`
`
`
`-10-
`
`
`
`Apple v. RED.COM
`Declaration of Graeme Nattress re POPR - IPR2019-010164 and IPR2019-01065
`
`
`19. Exhibit 2005, reproduced below, is a photo looking head-on into the
`
`front of Boris on the tripod. In the background of the photo, from left to right, is
`
`David Macintosh, Jarred Land and Ted Schilowitz. The lens of my camera is
`
`visible in the reflection of the lens on Boris. The reflection appears to show me
`
`adjusting the focus of my camera as Jarred Land simultaneously adjusts the focus
`
`on Boris.
`
`
`
`
`
`
`
`
`
`-11-
`
`
`
`Apple v. RED.COM
`Declaration of Graeme Nattress re POPR - IPR2019-010164 and IPR2019-01065
`
`
`20. Exhibit 2006, reproduced below, shows Ted Schilowitz posing next
`
`to a ColorChecker chart, also referred to as a Macbeth chart, while Boris is
`
`recording video of the shot. In the upper-right-hand foreground, an image of Ted
`
`Schilowitz can be seen in the monitor connected to Boris showing the live feed of
`
`the video.
`
`
`21. After recording test footage from Boris, we took the compressed raw
`
`Bayer-pattern video image data that Boris had saved to file, and played it back on a
`
`computer in decompressed and demosaiced form to examine the results. Exhibit
`
`
`
`-12-
`
`
`
`Apple v. RED.COM
`Declaration of Graeme Nattress re POPR - IPR2019-010164 and IPR2019-01065
`
`2007, reproduced below, shows Ted Schilowitz reviewing the output of the Boris
`
`test footage taken of himself.
`
`
`
`22. The results of the test footage were a resounding success. Visual
`
`inspection of the playback of the video files recorded by Boris confirmed that it
`
`had recorded compressed raw mosaiced video image data that had been outputted
`
`in visually lossless form with no visible compression artifacts after decompression
`
`and demosaicing.
`
`23. While the visually lossless nature of the Boris test footage was evident
`
`upon visual inspection alone, confirmation of its high resolution and frame rate
`
`was also confirmed by the computer program used to display the Boris video
`
`-13-
`
`
`
`Apple v. RED.COM
`Declaration of Graeme Nattress re POPR - IPR2019-010164 and IPR2019-01065
`
`footage. Although the output parameters of the test footage of Ted Schilowitz
`
`cannot be seen in the photo above, a photo of the output of test footage taken of
`
`David Macintosh provides a clearer image. That photo, Exhibit 2008, is
`
`reproduced below with a yellow circle annotation to indicate the output
`
`parameters.
`
`
`
`
`
`
`
`
`
`
`
`
`
`-14-
`
`
`
`Apple v. RED.COM
`Declaration of Graeme Nattress re POPR - IPR2019-010164 and IPR2019-01065
`
`
`24. Below is an enlargement of the output parameters contained in the
`
`annotated circle in Exhibit 2008.
`
`
`
`
`
`The output parameters read as follows:
`
`Format Settings:
`
`Format Custom: W: 4096 H: 2048
`
`Aspect Custom: Scale: 1.0000
`
`Framerate: 24.000
`
`Show: Full res. High.
`
`-15-
`
`
`
`Apple v. RED.COM
`Declaration of Graeme Nattress re POPR - IPR2019-010164 and IPR2019-01065
`
`At the top of the enlargement, the “.jim” file extension we used to denote
`
`compressed raw Bayer-pattern video image data created by REDCODE can be
`
`seen.
`
`25. These parameters report an output resolution of 4096 x 2048, which
`
`denotes 4K resolution. The framerate is reported as 24 frames per second. Based
`
`on these parameters, and my inspection of the visually lossless video shot by Boris,
`
`I knew by at least March 8, 2007 that our invention would work for its intended
`
`purpose. My understanding was then further confirmed later in March when Sir
`
`Peter Jackson successfully used Boris and its companion camera, Natasha, to shoot
`
`his “Crossing the Line” film.
`
`The April 2, 2007 Release Note for Boris and Natasha
`
`26. As part of my research and development of REDCODE, I instructed
`
`Wind River on a regular basis and conveyed how REDCODE needed to operate on
`
`the RED ONE cameras.
`
`27. Wind River would routinely provide documentation detailing the
`
`progress of the tasks we had assigned them. Attached as Exhibit 2010 is a true
`
`and correct copy of a Release Note, dated April 2, 2007, that Wind River provided
`
`to RED in connection with the Boris and Natasha cameras. This document was
`
`typical of the documentation that Wind River would generate on the RED ONE
`
`-16-
`
`
`
`Apple v. RED.COM
`Declaration of Graeme Nattress re POPR - IPR2019-010164 and IPR2019-01065
`
`project, and provide to RED on or near the date indicated on the cover. RED has
`
`maintained a copy of Exhibit 2010 since the RED ONE project.
`
`28. The cover of the April 2, 2007 Release Note states that it was
`
`“Prepared for Diamond.” Diamond was Wind River’s code name for RED. The
`
`cover also indicates that the document is a “Release Note” for “Sundance.”
`
`Sundance was Wind River’s code name for the RED ONE. Page four of the April
`
`2, 2007 Release Note indicates that the “released cameras are designated as
`
`‘Natasha’ and ‘Boris.’”
`
`29. Page five of the April 2, 2007 Release Note states, “Captured .jim
`
`files were verified by Diamond’s Graeme and David Macintosh to produce
`
`expected quality and the expected results from the new RED/BLUE pre-emphasis
`
`feature.” I am the Graeme referred to in this sentence. The “new RED/BLUE pre-
`
`emphasis feature” refers to an adjustment to the pre-emphasis functionality already
`
`running in REDCODE on Boris and Natasha. The “.jim” files are the compressed
`
`raw Bayer-pattern video image data produced by REDCODE running on Boris and
`
`Natasha.
`
`Huffman Compression
`
`30. As mentioned previously, the RED ONE cameras, including Boris and
`
`Natasha, relied upon JPEG2000 compression, which is a mathematically lossy
`
`compression technique. Huffman compression, as recited in the claims of the ’299
`
`-17-
`
`
`
`Apple v. RED.COM
`Declaration of Graeme Nattress re POPR - IPR2019-010164 and IPR2019-01065
`
`patent, refers to a lossless compression technique that has been well known for
`
`many decades.
`
`31. A lossless compression technique provides a lower compression ratio
`
`than lossy compression. While this difference impacts the write speed and
`
`memory on the hard drive, it does not lower the visual quality of the output upon
`
`decompression. Because the REDCODE on Boris and Natasha produced visually
`
`lossless output with a lossy compression technique, a person of ordinary skill in the
`
`art would know that RECODE would successfully produce visually lossless output
`
`using a lossless compression routine like Huffman compression. This result
`
`follows necessarily because a lossless compression technique removes any
`
`potentiality for compression artifacts caused by lossy compression.
`
`
`
`I declare 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.
`
`-18-
`
`
`
`Apple v. RED.COM
`Declaration of Graeme Nattress re POPR - IPR2019-010164 and IPR2019-01065
`
`Executed on August 15, 2019 at Acton, Ontario, Canada.
`
`30703838
`
`-19-
`
`