`
`1‘9 [Pf-10"]
`
`11111<
`WWWWWWIonem2969!:
`
`IN THE UNITED STATES PATENT AND TRADEMARK OFFICE
`REQUEST FOR FILING A PROVISIONAL APPLICATION FOR PATENT
`UNDER 37 CFR §1.53 (c)
`
`‘A
`
`INVENTOR(S)
`
`Levanon
`II
`‘ Lavi
`
`
`
`l 3 Nachal Besor St., Ramat Hasham, Israel
`
`21 Bar llan St., Raanana, Israel
`
`TITLE OF THE INVENTION
`
`OPTIMIZATION OF RENDERING MEMORY USAGE IN FAST QUALITY
`BUILD-UP TARGET IMAGE TRANSFER OVER LIMITED AND
`NARROWBAND COMMUNICATION NETWORKS
`
` :1 Signature:
`
`Direct all correspondence to Customer Number 23488.
`
`Gerald B. Rosenberg, Esq.
`NewTechLaw
`
`(Reg No.: 30,320)
`
`Telephone:
`Facsimile:
`
`650.325.2100
`650.325.2107
`
`WMWWWWWWW
`
`23488
`PATENT TRADEMARK OFFICE
`
`I 285 Hamilton Avenue, Suite 520
`Palo Alto, California 94301
`
`ENCLOSED APPLICATION PARTS (check all that apply)
`
`Specification
`
`No. of pages:
`
`Drawings
`
`Decla ration
`
`No. of sheets:
`
`I 1
`
`5
`
`Other: Return-Receigt Post Card.
`
`Small Entity Statement
`
`Power of Attorney
`
`Assignment and Cover Sheet
`
`METHOD OF PAYMENT OF FILING FEES FOR THIS PROVISIONAL APPLICATION FOR PATENT
`
`: Provisional Basic Filing Fee: $ 150.00 (Small Entity: $75.00)
`
`Filing Fee Amount: $ 150.00
`
`A check is enclosed to cover the Filing Fees.
`
`The Commissioner is hereby authorized charge Filing Fees or credit any
`overpayment to: Deposit Account Number: 50—0890.
`
`This invention was 991 made by or under contract with a US Government agency.
`
`US Government agency and Contract:
`
`Gerald B. Rosenberg
`Reg. No.: 30.320
`
`Date: December 26I 2000
`
`Application Docket No:
`
`FLVT3004
`
`__
`
`Express Mail Label No.2
`
`EL 661 534 291 US
`
`; Address To:
`
`Box Provisional A Iication Assistant Commissioner for Patents Washinton DC 20231
`
`gbr/flvb’3004.002 prov xmittal wpd
`
`10f17
`
`Microsoft Corp. Exhibit 1067
`
`1 of 17
`
`Microsoft Corp. Exhibit 1067
`
`
`
`OPTIMIZATION OF RENDERING MEMORY USAGE IN
`
`FAST QUALITY BUILD-UP TARGET IMAGE TRANSFER
`OVER LIMITED AND NARROWBAND
`
`COMMUNICATION NETWORKS
`
`Inventors:
`
`Isaac Levanon
`
`Yoni Lavi
`
`Background of the Invention
`
`The present invention is generally related to the delivery of high-resolution
`
`highly featured graphic images over limited and narrowband communications
`
`channels.
`
`Summary of the Invention
`
`The obiective is to display a two-dimensional pixel map, aI 6-Bit RGB color
`
`image in the preferred embodiments, of very large dimensions and permitting the
`
`viewing of the image from a dynamic three-dimensional viewpoint. Multiple such
`
`images are remotely hosted for on-demand selection and transfer to a client
`
`system for viewing.
`
`Attorney Docket No.: FLVT3004
`gbr/tlvt/3004.000.provnsional.wpd
`
`I 2/26/2000
`
`I
`
`2
`3
`
`4
`
`5 6 7 8 9
`
`IO
`
`I I_
`I2
`
`I3
`
`I4
`
`I5
`
`I 6
`
`I7
`
`I8
`
`I9
`
`20
`
`2I
`
`22
`
`23
`
`24
`
`25
`
`
`
`‘
`
`2 of 17
`
`Microsoft Corp. Exhibit 1067
`
`2 of 17
`
`Microsoft Corp. Exhibit 1067
`
`
`
`
`
`1
`
`2
`
`3
`
`4
`
`5
`
`6
`
`7
`
`8
`
`9
`
`10
`
`1 1
`
`12
`
`13 1
`
`14
`
`15
`
`16
`
`17
`
`18
`
`19
`
`20
`
`21
`
`22
`
`23
`
`24
`
`25
`
`-2-
`
`Images, as stored by the server, may individually range from gigabytes to
`
`multiple terabyte in total size. A correspondingly large server storage and
`
`processing system is contemplated. Conversely, client systems are contemplated
`
`to be conventional personal computer systems and, in particular, mobile, cellular,
`
`embedded, and handheld computer systems, such as personal digital assistants
`
`(PDAs) and internet-capable digital phones, with relatively limited to highly
`
`constrained network communications capabilities. For most wireless applications,
`
`conventional narrowband communications links have a bandwidth of less than
`
`approximately three kilobytes of data per second. Consequently, transmittal of
`
`entire images to a client system in reasonable time is infeasible as a practical
`
`matter.
`
`Overview:
`
`Description of the Invention
`
`For purposes of the present invention, each image (Figure 1) is at least
`
`logically defined in terms of multiple grids of image parcels with various levels of
`
`,
`
`resolutions (Figure 2) that are created through composition of information from
`
`all level of resolutions, and stored by the server to provide an image for transfer
`
`to a client system (Figure 3). Composed and separate static and dynamically
`
`created layers are transferred to client system in parcels in a program selectable
`
`order to optimize for fast quality build-up of the image presented to a user of the
`
`client system, particularly when the parcels are streamed over a narrowband
`
`communication link.
`
`The multiple layers of an image allow the selectivity to incorporate
`
`topographical, geographical, orientational, and other terrain and mapping
`
`Attorney Docket No.: FLVT3004
`gbr/flvt/3004.000.provisional.wpd
`
`12/26/2000
`
`3 0f 17
`
`Microsoft Corp. Exhibit 1067
`
`3 of 17
`
`Microsoft Corp. Exhibit 1067
`
`
`
`-3-
`
`related information into the image delivered. Other layers, such as geographic
`grids, graphical text overlays, and hyperlink selection areas, separately provided
`or composed, aid in the useful presentation and navigation of the image as
`presented by the client system and viewed by the user.
`
`Compositing of layers on the server enables the data transfer burden to be
`
`reduced, particularly in analysis of the requirements and capabilities of the client
`
`system and the connecting communications link. Separate transfer of layers to the
`client system allows the client system selectivity in managing and presentation of
`the data to the user.
`
`The system and methods of the present invention are designed to, on
`demand, select, process and immediately transfer data parcels to the client
`system, which immediately processes and displays a low-detail representation of
`the image requested by the client system. The system and methods immediately
`continue to select, process and sequentially transfer data parcels that, in turn, are
`processed and displayed by the client system to augment the presented image
`and thereby provide a continuously improving image to the user.
`
`Selection of the sequentially transferred data is, in part, dependent on the
`progressive translation of the three-dimensional viewpoint as dynamically
`modified on the client system during the transfer process. This achieves the
`above-stated objective while concurrently achieving a good rendering quality for
`continuous fly-over of the image as fast as possible, yet continuously building the
`image quality to the highest resolution of the image as stored by the server.
`To optimize image quality build-up over
`limited and narrowband
`
`1
`2
`3
`4
`
`5
`
`6
`
`7
`8
`9
`
`10
`11
`12
`13
`14
`15
`
`
`
` 16
`
`1 7
`18
`19
`20
`21
`22
`23
`
`24
`25
`
`is
`communication links, the target image, as requested by the client system,
`represented by multiple grids of 64x64 image pixels (Figure 4) with each grid
`
`Attorney Docket No.: FLVT3004
`gbr/flvt/3004.000.provi5|onal.wpd
`
`/
`
`12/26/2000
`
`4 of 17
`
`Microsoft Corp. Exhibit 1067
`
`4 of 17
`
`Microsoft Corp. Exhibit 1067
`
`
`
`-4_
`
`1
`
`2
`
`3
`
`4
`
`5
`
`6
`
`7
`
`8
`.9
`10
`
`1 1
`
`12
`
`.13
`i4
`15
`16
`1 7
`18
`
`19
`
`20
`
`21
`
`22
`
`23
`
`24 '
`
`25
`
`having some corresponding level of detail. That is, each grid is treated as a
`
`sparse data array that can be progressively revised to increase the resolution of
`
`the grid and thereby the level of detail presented by the grid. ‘ The reason for
`
`choosing the 64x64 pixel dimension is that, using current image compression
`
`algorithms, 0 16-bit 64x64 pixel array image can be presented as a 2KByte data
`
`parcel.
`
`In turn, this 2KByte parcel is the optimal size, subject to conventional
`
`protocol and overhead requirements, to be transmitted through a 3KByte per
`
`second narrowband transmission channel. Using a smaller image array, such as
`32x32, would create a 0.5KByte parcel, hence causing inefficiencies due to packet
`transmission overhead, given the nature of current wireless communications
`
`protocols.
`
`Image array dimensions are preferably powers of two so that they can be
`
`used in texture mapping efficiently. Each parcel, as received by the client system,
`is preferably immediately processed and incorporated into the presented image.
`To do so efficiently, according to the present invention, each data parcel
`is
`independently processable by the client system, which is enabled by the selection
`and server-side processing used to prepare a parcel fortransmission.
`In addition,
`each data parcel is sized appropriate to fit within the level-1 cache, or equivalent,
`
`of the client system processor, thereby enable the data processing intensive
`
`operations needed to process the data parcel to be performed without extended
`
`memory access delays.
`
`In the preferred embodiment of the present invention,
`
`data parcels are also processed for texture mapping and other image features,
`
`such as topographical detailing.
`
`Currently, with regard to conventional client systems, a la rger image array,
`
`such as 128x128, is too large to be fully placed within the level-1 cache of many
`
`
`
`
`
`
`
`
`Attorney Docket No.: FLVT3004
`gbr/flvt/3004.000.provisional.wpd
`
`12/26/2000
`
`5 of 17
`
`Microsoft Corp. Exhibit 1067
`
`5 of 17
`
`Microsoft Corp. Exhibit 1067
`
`
`
`1
`
`2
`
`3
`
`4
`
`5
`
`6
`
`7
`
`8
`
`9
`
`10
`
`1 1
`
`12
`
`13
`
`14
`
`15
`
`16
`
`17
`18
`19
`
`20
`
`21
`
`22
`
`23
`
`24
`'25
`
`
`
`3
`
`-5-
`
`of the smaller conventional current processors, such as used by personal digital
`
`assistants (PDAs) and cellular phones.
`
`Since access to cache memory is
`
`substantially faster than to RAM this will likely result in lower frame rate.
`
`Different and larger data parcel sizes may be optimal as transmission
`
`protocols and micro-architectures of the client computers change. For purposes
`
`above, the data content was a pixel array representing image data. Where the
`
`data parcel content is vector, text or other data that may subiect to different client
`
`system design factors, other parcel sizes may be used.
`
`ln the process implemented by the present invention, data parcels maybe
`
`selected for sequential transmission based on a prioritization of the importance
`
`of the data contained. The criteria of importance maybe defined as suitable for
`
`particular applications and may directly relate to the presentation of image
`
`quality, provision of a textual overlay of a low—quality image to quickly provide a
`
`navigational orientation, or the addition of topography information at a rate or
`
`timing different from the rate of image quality improvement. Thus, image data
`
`layers reflecting navigational cues,
`
`text overlays, and topography can be
`
`composed into data packets for transmission subject to prioritizations set by the
`server alone, based on the nature and type of the client system, and interactively
`influenced by the actions and commands provided by the user ofthe client system
`
`(Figure 5).
`
`'
`
`Progressive transmission of image parcels is performed in an iterative
`
`process involving selection of an image data grid within the target image of the
`
`client system, which is a portion of a potentially multi-Iayered source image stored
`
`by the server. The selection parameters are preferably dependent on the client
`navigation viewpoint, effective velocity, and height, and the effective level of detail
`
`Attorney Docket No.: FLVT3004
`gbr/flvt/3004.000.provisional.wpcl
`
`12/26/2000
`
`6 of 17
`
`Microsoft Corp. Exhibit 1067
`
`6 of 17
`
`Microsoft Corp. Exhibit 1067
`
`
`
`.....a
`
`2
`
`3
`
`4
`
`5
`
`6
`
`7
`
`8
`
`9
`
`10
`
`1 1
`
`12
`
`13
`
`14
`
`15
`
`16
`
`17
`18
`
`19
`
`20
`
`21
`
`22
`
`23
`
`24
`25
`
`
`
`-6-
`
`currently presented in each grid. Once a grid is selected, the server selects the
`
`‘ source data to be logically composed into the selected grid to complement the
`
`effective resolution of that grid, processing the grid data to produce the optimally
`
`sized size grid data parcels, and sequentially transmitting the parcels to the client
`
`system. Preferably, the detail of a grid array is sequentially enhanced by division
`
`of the grid into sub-grids related by a power of two (Figure 6). Thus, a given grid
`
`is preferably updated using four data parcels having twice the data resolution of
`
`the existing grid. Whatever number of parcels are used, each data parcel is
`
`rendered by the client system into the target image. Additional client system
`
`image data processing to provide texturing and three-dimensional representation
`
`of the data may be performed as part of the parcel rendering and integration into
`
`the target image.
`
`Image Parcel Download Seguence:
`
`The server of the present invention supports the download of parcel data
`
`to a client system by providing data parcels in response to network requests
`
`originated by client systems. Each requested data parcel is identified within a grid
`coordinate system relative to an image stored by the server.
`
`A client system implementing the process of the present invention is
`
`responsible for identifying and requesting parcel data, then rendering the parcel
`
`data into the target image at the correct location. The client system is also
`
`responsible for managing navigational and other interaction with the user.
`
`ln
`
`identifying the parcel data to be requested, the client system operates to select
`
`grids within the coordinate system, corresponding to portions of the target image,
`for which to request a corresponding data parcel. The requests are issued over
`
`Attorney Docket No.: FLVT3004
`gbr/flvt/3004.000.provisional .wpd
`
`12/26/2000
`
`7 0f 17
`
`Microsoft Corp. Exhibit 1067
`
`7 of 17
`
`Microsoft Corp. Exhibit 1067
`
`
`
`-7-
`
`the network to the server and rendering performed asynchronously as data
`
`parcels are received. The order of data parcel requests is defined as a sequence
`
`that will provide for the optimal build-up of the target image as presented to the
`
`user. The rate of optimal build up of the target image is dependent on the nature
`
`>
`
`of the target image requested, such as the supported parcel size and depth of the
`
`target image that can be rendered by the client system.
`
`The client identifies and requests the download of data parcels in the
`
`process as follows. Denote the target image as IO and its size in pixels as (X, Y).
`
`Let N be the smallest power of 2 that is equal or greater than max {X,Y}.
`
`1
`
`2 3 4 5
`
`6
`
`7
`
`8
`
`9
`
`
`
`1O
`
`11
`
`Construct the grid of 64x64 pixel grid-images 101i,i that together compose the
`
`target image 10. The rectangle [64i,64i + 64] x [64),64 i + 64] of I0 is mapped
`
`12
`
`13
`
`14
`
` 15
`
`~
`
`16
`
`17
`18
`
`19
`
`2O
`
`21
`
`22
`
`23
`
`24
`
`25
`
`to lo“.
`
`In order to view a large portion of the image, the target image, without
`
`downloading the substantial bulk of the target image, mip-maps of 10 are created,
`
`representing a collection of images to be used as surface textures when rendering
`
`a two-dimensional representation of a three-dimensional scene, and which are
`
`defined recursively as:
`
`lk+1 (i,i) = avg(|k (2i,2i), lk(2i + 1,2i), Ik(2i,2i + 1), lk(2i + 1,2i + 1))
`
`Such mip-maps are created up to IM,M = log2 (N) - 6. At this point, |M is
`
`a 64x64 image containing the entire area of the original image, hence no further
`
`mip-mapping is required.
`
`The methods of the present invention then proceed by constructing the
`
`respective grids or cells (lklili) for each mip-map. Each nonempty image cell lkli,i
`
`Attorney Docket No.: FLVT3004
`gbr/flvt/3004.000.provisiona|.wpd
`
`12/26/2000
`
`8 of 17
`
`Microsoft Corp. Exhibit 1067
`
`8 of 17
`
`Microsoft Corp. Exhibit 1067
`
`
`
`-8-
`
`CIJ\IO\U‘I-I>~OJM—-'
`
`9
`
`10
`
`1 1
`
`12
`
`13
`
`14
`
`now may be downloaded. Larger values of k cover more area within the original
`
`image but provide lower detail on that area. The task at hand is now to
`
`determine, given the viewing frustum and the list of previously downloaded image
`
`cells lklili, downloading which grids will improve the quality of the display as fast
`
`as possible, considering the download rate as fixed. The scheme used to
`
`implemented the downloading sequence of these cells is by constructing a tree,
`
`starting from IN“),O and expanding a quadtree towards the lower mip-map levels.
`
`(Quadtrees are data structures in which each node can have up to four child
`
`nodes. As each 64x64 pixel image in the grid 1k has exactly four matching 64x64
`
`pixel images on the grid 1k_1 covering the same area, the data structure is built
`
`accordingly.)
`
`For every frame that is rendered , begin with the cell that covers the area
`
`of the entire original image, lN_6IOIO. For each cell under consideration, compute
`
`the principle mip-map level that should be used to draw it.
`
`If it is lower than the
`
`. 15
`
`mip-map level of the cell, subdivide the cell
`
`to four smaller cells and use
`
`16
`
`17
`
`18
`
`19
`
`20
`
`21
`
`22
`
`23
`
`24
`
`25
`
`recursion.
`
`lf this operation attempts to draw over areas that do not yet have
`
`image cells at a low enough mip-map level to use with them, the recursion stops.
`
`If the principle mip-mop level is equal or higher than the level of the cell,
`
`then the cell is rendered using the cell of the principle mip-map level, which is the
`
`parent of that cell in the Quad-tree, at the appropriate level. Then download the
`
`cells in which the difference between the principle mip-map level to the mip-map
`
`level of the image cell actually used is the highest. Downloading is asynchronous;
`
`the renderer maintains a priority queue of download requests, and separate
`
`threads are downloading images. Whenever a download is complete, another
`
`download is initiated immediately, based on the currently highest-priority request.
`
`
`
`
`
`Attorney Docket No.: FLVT3004
`gbr/flvt/3004.000.provisional.wpd
`
`12/2 6/2000
`
`9 of 17
`
`Microsoft Corp. Exhibit 1067 _
`
`9 of 17
`
`Microsoft Corp. Exhibit 1067
`
`
`
`-9-
`
`The principle mip-map level of an image cell is determined by the screen
`
`resolution, FOV (tield ot view) angle, the angle tormed between the image's plane
`
`normal and the line connecting between the camera and the position within the
`
`cell that is closest to the camera, and a few other factors. The equation, which
`
`uses the above information, approximates the general mip-mapping level
`
`equation:
`
`| = maxlO, log4 (T/S))
`
`in which 8 is the surface of the cell as displayed on the screen during rendering
`
`(in pixels), and T is the surface of the cell within the texture being mapped (iii
`pixels).
`
`When rendering a cell of the grid lk,
`
`T = N22k
`
`and
`
`5 = xycos(a)ctg2(0.5FOV)t2 T / 22
`
`where x is the display's x-resolution, y is the display's y-resolution, FOV is the field-
`
`ot—view angle, a is the angle between the image's plane normal and the line
`
`connecting the viewpoint and the point in the cell of shortest distance to it, t is the
`
`length of the square each pixel in the original image is assigned to in 3D, and z
`
`is the height of the camera over the image's plane.
`
`This arrives at the equation:
`
`l = log4 (z2 /(xycos(a)ctg2(0.5FOV)t2))
`
`Attorney Docket No.: FLVT3004
`gbr/tlvt/3004.000.provisional .wpd
`
`12/26/2000
`
`l
`
`2
`
`3
`
`4
`
`5
`
`6
`
`7 8
`
`9
`
`10
`
`it
`12
`
`13
`
`14
`
`15
`16
`
`l7
`18
`
`19
`
`20
`
`2]
`
`22
`
`23
`
`24
`
`25
`
`10 0f 17
`
`Microsoft Corp. Exhibit 1067
`
`
`
`
`
`
`__.
`
`10 of 17
`
`Microsoft Corp. Exhibit 1067
`
`
`
`-10-
`
`l = max(0, min(l, M))
`
`For example, using a 64x64 target grid display to render the image from a view
`
`of height N with FOV angle of 90 degrees, with the length of each pixel in space
`
`being one, the entire target image can be fitted precisely to the display as
`
`demonstrated by:
`
`I = Iog4(N2/(642 -1 -1 ~12» = M
`
`Rendering Memory Garbage Collection Process:
`
`The memory amount used by the above image parcel data selection and
`
`rendering process can be,
`
`in general, extremely large. An efficient form of
`
`garbage collection is
`
`required to prevent out-of-memory conditions while
`
`minimally impacting the performance requirements of the client system.
`
`Particularly on low performance and memory constrained client systems, such as
`
`PDAs and cellphone, the garbage collection process needs to be efficient both in
`terms of the processor overhead as well as in the memory requirements of the
`garbage collection control structures.
`
`In the present invention, a texture cache is used as the basis for garbage
`
`collection. The texture cache includes a buffer with a pre—determined size that is
`
`sufficient to store a defined number of image data parcels. Every rendering of a
`
`cell places the resulting image data and the image data of all of its parent cells
`
`in a quad-tree structure stored sequentially upward in the buffer. Each time a new
`
`image is to be inserted into the texture cache buffer and there is insufficient free
`
`space, the image data for the quad-tree structure logically positioned at the
`
`Attorney Docket No.: FLVT3004
`gbr/flvt/3004.000.provisional.wpd
`
`12/26/2000
`
`1
`
`2 3
`
`4
`
`5
`
`6
`
`_ 7
`
`8
`9
`
`1O
`
`1 1
`
`12
`
`13
`
`14
`
`15
`
`16
`17
`18
`
`19
`
`20
`
`21
`
`22
`
`23
`
`24
`
`25
`
`
`
`11 0f 17
`
`Microsoft Corp. Exhibit 1067
`
`11 of 17
`
`Microsoft Corp. Exhibit 1067
`
`
`
`-11-
`
`1
`
`2
`
`3
`
`4
`
`bottom of the butter is removed and the corresponding buffer memory reclaimed
`
`as free buffer memory. The optimal size of the texture cache may not be
`
`determinable analytically, but is easily subiect to empirical analysis to identify an
`
`optimal size.
`
`
`
`Attorney Docket No.: FLVT3004
`gbr/tlvt/3004.000.provisional.wpd
`
`1 2/26/2000
`
`12 0f 17
`
`Microsoft Corp. Exhibit 1067
`
`12 of 17
`
`Microsoft Corp. Exhibit 1067
`
`
`
`511/.”
`
`
`
`2meU0:E__:3
`
`926355::
`
`mEmEmEamwcm0N
`
`
`
`t3“Eta>flEQEQEU
`
`$53580
`
`U20
`
`m>=omam$m
`
`5m5:50m
`
`,6mmmhmmu
`
`Eonwmt
`
`25225am
`
`255.58.
`
`13 0f17
`
`Microsoft Corp. Exhibit 1067
`
`13 of 17
`
`Microsoft Corp. Exhibit 1067
`
`
`
`
`
`
`
`
`FIG.2
`
`_ 14 0f 17
`
`Microsoft Corp. Exhibit 1067
`
`14 of 17
`
`Microsoft Corp. Exhibit 1067
`
`
`
`
`
`
`
`FIG. 4
`
`15 0f 17
`
`Microsoft Corp. Exhibit 1067
`
`15 of 17
`
`Microsoft Corp. Exhibit 1067
`
`
`
`
`
`
`FIGS
`
`16 0f 17
`
`Microsoft Corp. Exhibit 1067
`
`16 of 17
`
`Microsoft Corp. Exhibit 1067
`
`
`
`
`
`64
`
`64
`
`64
`
`
`
`17 0f 17
`
`Microsoft Corp. Exhibit 1067
`
`17 of 17
`
`Microsoft Corp. Exhibit 1067
`
`