Exhibit 2007
`Bradium Technologies LLC - patent owner
`Microsoft Corporation - petitioner
`IPR2016-01897
`1
`
`

`

`\OO0\lO\U'I-I>-CaOl\J—I
`
`
`MULTIPLE PARALLEL DOWNLOAD OF TARGET
`IMAGE PARCELS STREAMED OVER LIMITED AND
`NARROWBAND COMMUNICATIONS CHANNELS
`
`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 lrivention
`
`The objective is to display a two—dimensional pixel map, a I 6-Bit RGB color
`
`image in the preferred embodiments, of very large dimensions and permitting the
`
`viewing ofthe 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.
`
`Images, as stored by the server, may individually range from gigabytes to
`
`multiple terabyte in total size. A correspondingly large server storage and
`
`L Attorney Docket No.: FLVT3005
`gbr/flvt/3005.000.provisionaI.wpd
`
`I2/26/2000
`
`2
`
`

`

`
`
`-2-
`
`m\lO~U‘I-l>OOI\3—-'
`
`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 mostwireless 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 l) 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
`
`23
`
`24
`
`25
`
`topographical, geographical, orientational, and other terrain and mapping
`
`related information into the image delivered. Other layers, such as geographic
`
`grids, graphical text overlays, and hyperlink selection areas, separately provided
`
`Attorney Docket No.: FLVT3005
`gbr/flvt/3005.000.provisional.wpd
`
`l2/26/2000
`
`3
`
`

`

`——l
`
`o~ooo\ro~uw.i>'oor\>
`
`2|
`
`.1]
`
`——I
`
`.._|
`
`Ix)
`
`_.| 0.)
`
`—I -IA
`
`._| U1
`
`._l O
`
`.._a \l
`
`5 (X)
`
`._| ‘O
`
`l\’J O
`
`I\) ._I
`
`t\) (O
`
`Ix) 00
`
`IO J5-
`
`N) U‘!
`
`
`
`_3_
`
`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
`
`communication links, the target image, as requested by the client system,
`
`is
`
`represented by multiple grids of 64x64 image pixels (Figure 4) with each grid
`
`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
`
`Attorney Docket No.: FLVT3005
`gbr/flvt/3005.000.provis|onal.wpd
`
`l2/26/2000
`
`4
`
`

`

`-4-
`
`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, a 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 O.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-l 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 larger image array,
`
`such as l28xl 28, is too large to be fully placed within the level-1 cache of many
`
`of the smaller conventional current processors, such as used by personal digital
`
`l 2 3 4 5 6 7 8
`
`Attorney Docket No.2 FLVT3005
`gbr/tlvt/3005.000.provisional.wpd
`
`"I2/26/2000
`
`5
`
`

`

`._.a
`
`assistants (PDAs) and cellular phones.
`
`Since access to cache memory is
`
`-5-
`
`©‘0OO\lO\U‘l-hCaJI\7
`
`
`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.
`
`In 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 Cl
`
`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 of the 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-layered 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
`
`currently presented in each grid. Once a grid is selected, the server selects the
`
`Attorney Docket No.: FLVT3005
`gbr/flvl/3005.000.provisional.wpd
`
`1 2/2 6/2 000
`
`6
`
`

`

`-6-
`
`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.
`
`In
`
`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
`
`the network to the server and rendering performed asynchronously as data
`
`Attorney Docket No.: FLVT3005
`gbr/flvt/3005.000.provisional.wpcl
`
`T2/26/2000
`
`l 2 3 4 5 6 7 8
`
`7
`
`

`

`-7-
`
`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 identities and requests the download of data parcels in the
`
`process as follows. Denote the target image as lo 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}.
`
`Construct the grid of 64x64 pixel grid-images I0,“ that together compose the
`
`target image IO. The rectangle [64i,64i + 64] x [64i,64 i + 64] of I0 is mapped
`
`to lo“.
`
`In order to view a large portion of the image, the target image, without
`
`downloading the substantial bulk otthetarget image, mip-maps of to 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:
`
`|k+] (i,i) = avg(lk (2i,2i), lk(2i + l,2i), lk(2i,2i + 1), |k(2i + 1,2] + 1))
`
`Such mip-maps are created up to lM,M = log2 (N) - 6. At this point, IM 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 (lm) for each mip-map. Each nonempty image cell IW
`
`now may be downloaded. Larger values at k cover more area within the original
`
`Attorney Docket No.: FLVT3005
`gbr/flvt/3005.000.provisional.wpd
`
`12/26/2000
`
`l 2 3 4 5 6 7 8
`
`8
`
`

`

`-3-
`
`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 lm, 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 lN_6’o,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 lk has exactly four matching 64x64
`
`pixel images on the grid 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, |N_6,O,0. For each cell under consideration, compute
`
`the principle mip-map level that should be used to draw it.
`
`If it is lower than the
`
`mip-map level of the cell, subdivide the cell
`
`to four smaller cells and use
`
`recursion.
`
`If 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-map 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.: FLVT3005
`gbr/flvt/3005.000.provisioncil.wpd
`
`‘I2/26/2000
`
`
`
`9
`
`

`

`-9-
`
`The principle mip—map level of an image cell is determined by the screen
`
`resolution, FOV (field of view) angle, the angle formed 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:
`
`I = max(O, log4 (T/5))
`
`in which 5 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 (in
`
`pixels).
`
`When rendering a cell of the grid lk,
`
`T = N22"‘
`
`and
`
`S = xycos(a)ctg2(O.5FOV)t2 T / 22
`
`CD\OOO\lO\U‘I-h0ol\)#
`
`
`where x is the display's x-resolution, y is the display's y-resolution, FOV is the field--
`
`of-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 2:
`
`is the height of the camera over the image's plane.
`
`This arrives at the equation:
`
`I = log, (22 /(xycos(a)ctg2(O.5FOV)t2))
`
`Attorney Docket No.: FLVT3005
`gbr/flvt/3005.000.provisional.wpd
`
`12/26/2000
`
`10
`
`10
`
`

`

`-10-
`
`l = max(O, 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 = log, (N2/(642 '1 '1 '12)) = M
`
`
`
`ti.
`
`lmage Data Parcel Transfer for Multiple Target lmage Layers:
`
`The ordered download process of the present invention can be extended
`
`to support the concurrent download of multiple target image layers for separate
`
`or composited rendering by the client system. For example, one such layer may
`
`be a geographic map and a second layer may be a graphic-based image overlay
`
`providing textual and graphical navigational cues. Both will be viewed by the user
`
`from the same effective viewpoint and reflect navigational translations ofthe point
`
`of view over the map. Provided that both layers share the same coordinate
`
`system, data parcels representing the corresponding grid portions of each layer
`
`can be requested concurrently by the client system in execution of the process
`
`described above. That is, each layer image will be composed at each cell layer
`
`of corresponding mip-ma ps. The client system process of receiving and rendering
`
`data parcels is extended to distinguish between the data parcels as received for
`
`the different
`
`layers and, as appropriate, provides for the [oint or separate
`
`rendering of the parcel data.
`
`Attorney Docket No.: FLVT3005
`gbr/flvf/3005.000.provisional.wpd
`
`12/26/2000
`
`11
`
`11
`
`

`

`-11-
`
`CD\0OO\lO\U'I-l>0Ol\')—’
`
`Multiple Parallel Download of Target Image Parcels:
`
`The above concept can be extended to allow the viewing of multiple
`
`images from the same viewpoint, assuming that they are tiled (i.e. placed on CI
`
`grid), where each image within the grid contains mip-maps as previously
`
`described. The application tracks which of the images are currently viewed, and
`
`simultaneously performs downloading and rendering for all of them.
`
`For a given layer, some of the cells within the grid may not necessarily
`
`contain images, and can be ignored. This allows for handling even larger-scale
`
`images, because they can be first split into smaller images and then, potentially
`
`in parallel, compiled to produce the mip-map grid storage tiles required for each
`
`layer.
`
`
`
`ti».
`
`
`
`Attorney Docket No.: FLVT3005
`gbr/flvt/3005.000.provisional.wpd
`
`12/26/2000
`
`12
`
`12
`
`

`

`
`
`:6$5.:6$5.
`
`
`
`AA
`
`
`
`ff‘ff‘
`
`
`
`L_».~».:L_».~».:
`
`
`
`Em.mn_w.9__mu£..FEm.mn_w.9__mu£..F
`
`
`
`x£$m§§.,x£$m§§.,
`
`
`
`58...3.w>m__aoEn_E058...3.w>m__aoEn_E0
`
`
`
`
`
`
`
`mEmEm_momccmGmmEmEm_momccmGm
`
`
`
` S.U.... S.U....
`
`
`
`
`
`ms._§_E__::m>_.u2_m.mmms._§_E__::m>_.u2_m.mm
`
`
`
`
`
`
`
`
`
`.0.0
`
`
`
`
`
`
`
`m_ox_aum:E.__::x_m_.:_>>onm_ox_aum:E.__::x_m_.:_>>on
`
`M.MEonmm:%s“.0mwflmmcM.MEonmm:%s“.0mwflmmc
`
`
`
`m_m:_::m._.mm_5os_5amm_m:_::m._.mm_5os_5am 3%$.K,...\\2,5m,3%$.K,...\\2,5m,.,Nwo_%§m8wW.9.,Nwo_%§m8wW.9
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`1313
`
`13
`
`
`
`
`
`

`

`
`
`
`
`
`
`FIG.2FIG.2
`
`
`
`1414
`
`14
`
`

`

`
`
`
`
`
`
`iv"3i"::::::‘.:s§Z..)Zi£2.15::M‘::::§'siv"3i"::::::‘.:s§Z..)Zi£2.15::M‘::::§'s
`
`
`
`
`
`
`
`
`
`
`
`6464
`
`
`
`6464
`
`
`
`6464
`
`
`
`6464
`
`
`
`6464
`
`
`
`6464
`
`
`
`6464
`
`
`
`6464
`
`
`
`H6. 4H6. 4
`
`
`
`1515
`
`15
`
`

`

`
`
`......w......w
`
`
`
`1616
`
`16
`
`

`

`
`
`my65my65
`
`
`
`1717
`
`17
`
`