JPS5998275A - Method of mapping image on observation plane and circuit equipment for image address-memory address mapping of biaxial square array - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

DETAILED DESCRIPTION OF THE INVENTION [Technical Field of the Invention] The present invention relates generally to computer-controlled imaging devices and, more particularly, to computer-controlled imaging devices and, more particularly, to computer-controlled imaging devices and, more particularly, to the field of computer-controlled imaging devices. The present invention relates to a digital video processing apparatus having the ability to assemble and compose a sequential stream of scenes from a video library for display at processing speeds sufficient to permit real-time or near-real-time analysis of images.

One example of the many possible applications of such a device is in the field of vehicle simulation, such as seven-light simulation of aircraft. In such systems, a visibility subsystem within an aircraft flight simulation device receives flight data from a flight simulation computer and terrain data from a defined or "gaming area" database. A data processor within the visibility simulation device combines the flight data and terrain data to generate a simulated visual display for display to an observer in the cockpit of the aircraft.

For example, the visibility system of a vehicle simulator, such as a simulator for a helicopter, includes a "view through a window of the simulated surroundings," and a means to guide the "vehicle" in any desired direction relative to such surroundings. control. The term "view seen through the window" of this device is, for example, approximately 60 to 260 Krr12 (approximately 25 to 100 Krr12).
refers to a display, usually in the form of a video, of a simulated environment that corresponds to terrain covering a large area (200 square miles). This simulated surrounding situation will be referred to herein as a defined area or gaming area.

Operation of the vehicle controller directs the vehicle within, around, and through the gaming area. - What determines what is visible through the window, i.e. the video display, is
This is the device's response to the vehicle controller. What we see through the "window" is called the field of view, or FOV.

[Prior art]

One conventional device, known as a computer-generated imagery (CGI) device, utilizes a computer device to generate images from a database that can display images. Objects and grounds for constructing possible scenes are derived from a purely mechanical model, which is stored in the form of points delimiting objects and surfaces.

The strength of CGI lies in its surface appearance. A real ground or artificial surface can be measured to obtain the altitude at a specified point, usually the intersection of a uniform grid. By linking the heights of the specimens, the ground can be reconstructed by computer. In addition to realistic ground representation, CGI
also controls the placement of objects on the ground. Since altitude data is usually given on a uniform grid,
The placement of other objects can be specified on this same grid. Typical objects such as trees, rocks, shrubs, houses, roads, etc. may have defined positions in the grid system of the database.

Accurate lighting and perspective are also major contributions from CGI. Accurate illumination is achieved by finding the ground normal for each displayed pixel. This normal is used along with the line of sight, the normal from the light source plus the ambient brightness, to calculate the brightness for the pixel. Accurate perspective is achieved because the distance from the observation point to each surface point is known. This distance is an important variable in perspective transformation.

The weakness of this CGI is that it lacks realism.

Objects can be precisely positioned, accurately illuminated,
Although the object can be viewed correctly, the object itself cannot be realistically displayed. The current state of the art in CGI object display is such that objects are displayed in a very linear manner. Certain scenic elements, such as barren terrain, sand and clouds,
Can be displayed more realistically than complex objects such as trees, grass, or detailed artifacts. Such complex objects simply lack realism.

Another video generator is called a "computer composite image" or C8I. C8I technology also generates images such as video displayable images from a database, but the objects and ground stored in the database are
Rather than a mathematical model of those objects in CGI, they are displayed as real-world electromagnetic media images of the objects and the ground.

Thus, whereas CGI uses a computer to generate images from a purely mathematical database, C8I uses a computer to fit objects into a scene based on stored images of reality. . Although CGI provides great control over the scenes that are constructed and displayed to interact with the surrounding environment, it is less reliable and therefore less realistic in the scenes displayed. C8I
is the exact opposite. Fidelity is extremely high, but control over the structure of the scene is limited.

C8I's strength lies in the use of real-world images, such as photographs, in the scene. Currently available video equipment can easily handle photographic data. Individual photographs, such as sashiko characters, can be recorded on a video disc, and their recall can be controlled by an indexing device, just as is the case for digital data stored on a magnetic disc. Furthermore, the reliability of the image is reliable and the output image is the same as the input and stored image.

The weakness of the C8I is that its view is limited to the "camera" shooting point. In other words, it is not possible to dynamically navigate a scene unless a series of photographs are taken of the scene from beginning to end. For any gaming area of reasonable size, the number of sequence shots from beginning to end is prohibitively large.

[Summary of the invention]

According to the present invention, the CGI device is newly developed C8.
In combination with engineering technology, a ``computer-generated composite figure'' or CGSI device can be constructed. Here, the invention involves combining the best of two technologies, CGI and CS engineering, to construct CGS I. The scene is constructed by placing individual, usually detailed, objects in high fidelity (C8I) on a specified ground or background generated by CGI or C8I. CGSI sights can be constructed in much the same way as ω-engine sights, with ground elevations and object positions placed on a uniform grid. Individual objects used in the view are adjusted for perspective, and location and transformations, including dimensions, positions, rotations, torsions, and brightness, are performed for each image as required. The ground may be a CGI composition, or it may be a series of C8I ground inlays. - A scene is usually constructed by placing objects starting from the object furthest from the observation point, that is, the scene recognition device, and placing the objects at the point where the nearest object is placed. CGSI can be constructed from features from any portion of the electromagnetic spectrum, including the visible spectrum, IR spectrum, MMw spectrum, radar spectrum, etc.

Accordingly, it is an object of the present invention to provide a new and improved computer-generated modeling system that includes the use of real world images in a database.

Hereinafter, the present invention will be described in detail with reference to the drawings.

〔Example〕

Photo I shows the battlefield seen from the air. The battlefield can be called a gaming area or a defined area, and within practical constraints is approximately 60 to 260 km” (approximately 25
~1 o, o square miles) areas are common.

For example, if the video display visualization device of the present invention is used to simulate the operation of a helicopter,
A simulated gaming area as shown in Photo I can be selected as the environment in which to fly the helicopter. The helicopter simulator's visualization system provides a continuous "through-the-window" view of the gaming area. The display is a video representation of the sequence of sights seen by the pilot within the gaming area, corresponding to the position and altitude of the helicopter relative to the gaming area. The helicopter simulator incorporates a control device for freely guiding or flying the helicopter in any direction within the gaming area, around the gaming area, and through the gaming area. The simulator is responsive to its controller to determine what is visible within the "window" of the video display.

Photo II shows a scene somewhere within the gaming area of Photo I, probably in the shade of a grove of trees. The scene was challenged by the time on a video display that mimics the windows inside a helicopter simulator's training room. By the continuous operation of the control device by the trainee, the movement of the helicopter within the gaming area is determined, and the scene shown on the video display follows each instantaneous position of the helicopter.

Photos IIA, IIB are actual copies of captured photographs of displayed video display scenes within the gaming area, generated in accordance with the principles of the present invention. From these photographs, note the realism of the details of the objects, even though the images of the scene were obtained through several processing steps that had the effect of blurring the details. Furthermore, the smooth transition 9 from the object to the background indicates that the scene is represented without a cartoonish expression.

FIG. 1 is a block diagram of the CGS I device.

Pictures III-XI illustrate the steps in the construction of a typical C3GI device ending at block 12 of FIG. These photographs were also taken of the displayed images.

The composition of a Cps I scene usually begins by placing the ground, water, and sky first. After that, place the large object and the small object. Objects include trees, rocks, houses, shrubs, roads, lights, vehicles, helicopters, winged aircraft, animals, and women. Finally, you can add special effects if you wish, such as smoke, clouds, shadows, etc. To illustrate how CGS I works, the construction of sample scenes will be described with reference to photographs E-XI.

Starting from photo III, a section of sky is added on top of the background at a certain distance. By dividing the sky, peaks and valleys can form the horizon shown in the photo. In this example, the sky is divided into five parts. In general, the lower edge of the segment need not be straight, but may be gently undulating,
It can also be curved or jagged to simulate steep hills or mountains. A description of how to twist individual sections based on elevation angles based on the lowest and highest data and observation points can be found at
More on this later.

In Photo IV, rough ground is added to the section to form the foreground or foothills of the mountain. The untouched area between the base of the mountain and the sky is bounded by mountains in the distant background. In subsequent photographs, a stored textured ground plane is added to the scene, distorted to fit the screen coordinates of the ground polygon. The brightness of each ground can be varied depending on distance and other desired parameters.

Photo V shows the planned road section.

For that road portion, the road representation in the database ground library is warped to fit the screen coordinates. The ground library may include various road surfaces and other special ground surfaces such as streams and ponds.

Photo VI shows an example of a relatively small two-dimensional (2D) object that has been planned. Those objects occupy less than a predetermined percentage of the total screen. In one embodiment, objects occupying less than 1/16 of the area of the scene are represented in 2D. This is because, in many applications, relatively small natural objects such as trees, bushes, and rocks can be represented from one side, i.e. as two-dimensional objects, with little loss of realism. . Objects that cannot be represented from one side, such as large buildings or objects of special interest such as tanks, ships, etc., are called three-dimensional objects (3D) and are represented as such. Relatively small 2D objects can be processed with less expensive processing hardware/software than that required to process 3D objects and surfaces. While flying through a scene, a 2D object can be transferred to a 3D processor when it occupies more than a preselected percentage of the area of the scene.

Photo VII shows the tank in multiple views or 3D. Multiple views of the tank are retracted and the correct view is used to compose the scene based on the tank's path, altitude and observer's perspective. The tank may be mobile and may be very large.

Photo VIII shows a house that is an example of a multi-surface or 3D building. The house is separated into several surface parts, several roof parts (one if both sides are the same), two ends and two sides.

The individual surfaces of the house can be twisted from the normalized view to form a perspective directed by screen coordinates, and then tied together.

Photo IX shows a large 2D object. Those objects can occupy more than a predetermined amount of the area of the scene. If desired, the objects can be expanded so that they are larger than the entire surface of the screen.

Photo X illustrates special effects techniques used for transparent media including clouds, flakes, smoke, and shadows. The mask controls the transparency function and the second input word controls the brightness and color.

Photo XI shows the complete CGSI view.

The scene appears somewhere in the gaming area shown in Photo I.

The block diagram of FIG. 1 will now be referred to. Each item, from the database structure of the block diagram to the incorporation of special effects, will be briefly explained below and detailed later.

Database Structure The database contains two very different types of data, each related to an object library and a gaming region. The object library software generates features and stores those features on optical disks with high fidelity. Gaming region hardware is used to load locations of objects, surfaces, special effects, and more.

The flexibility of object libraries is almost limitless. The object library may include images of objects, images of the ground, and special effects transmission masks. They can be represented by one of a number of electromagnetic radiation spectrum bands. As a result, it is possible to simulate not only the visible range but also human power/output based on the detected IR spectrum, ■den spectrum, radar spectrum, etc.

The object library can also contain a mixture of 2D and 3D images. Those images can represent day/night conditions, daily conditions. The visible object library typically contains photographic objects. When constructing a high-fidelity object from an object library, objects stored in the library such as individual real-world elements, highly accurate models, paintings, photographs, etc. are recalled to create a "near-perfect" object. form an image.

It restores edges, separates objects from the background, corrects brightness and color, generates realistic colors, positions objects from system reference points, and generates high-fidelity CGI objects. This is done by generating graphic data, ie, a light source. A landing point and altitude reference point are also added. "Near perfect" objects, terrain and special effects are stored on fast access and high data rate media. [The term near-perfect means high fidelity with respect to the quality of the input video.

The gaming region database provides the information necessary to place the contents of the object library, terrain, and special effects onto a grid or gaming region. Objects can be placed by an operator or randomly by a computer.

Objects in the library can be held still or moved. The output of this function determines the content of the scene.

Vehicle Simulation Calculations Vehicle simulation calculations are based on a mathematical model of the vehicle and control inputs to determine the location and viewing orientation of the primary vehicle's visual or sensor devices.

Calculations can also be performed for a second vehicle based on the vehicle model and the selected route. This output determines the observer's location.

Communication Subsystem Of course, the input/output or Ilo of the vehicle simulation device and the Ilo of the CGSI device must interface in an efficient manner. A communications subsystem can be a bidirectional link and buffer that interfaces two devices.

Field of View and Coordinate Transformation Calculation The FOV processor determines the presence of objects, ground, and special effects in the scene being constructed. The output of the transformation matrix (V) transforms real world coordinates to screen coordinates.

This data from the transformation matrix allows rapid testing and determination of any part of the object, ground, and special effects that are all present in the scene. To avoid testing whether all objects exist in the database, a "smart" algorithm tests only objects and surfaces that are close to the scene. The FOV processing device generates a table of objects within the FOV and 1. A table of channel assignments for those objects/grounds or special effects is maintained. The function of the FOV computer is to be able to determine what the observer can see. The object, ground, and special effects controllers "fan out" and process the control functions generated during FOv calculations. The processed control functions are sent to the object/ground/special effects channel. The main functions performed by the control unit are converting gaming area coordinates to screen coordinates, processing distance data from the vehicle controlled by the operator to each object in the FOv, and identifying distances and objects. Instruct the object library base to determine the brightness of each object based on the image data and to search for accurate image data.

The function of the controller is to "fan out" the FOv data and generate precise control data for the scene.

Library of Objects, Grounds, and Special Effects This library stores images used in scene composition.

A controller directs selected images to be sent to the processing channel. The library's only function is to store images and provide the correct images on command.

Processing Channels for Objects, Grounds, and Special Effects Individual processing channels or "pipeline processors" typically process 91 large items (objects, grounds, or special effects) at a time per channel. A processing channel may have the ability to process multiple smaller items in parallel. All processing channels operate in the same way for each such item1, since it is the nature of the item that specifies the channel's functionality.

In one embodiment, each processing channel modifies one large item or sixteen small items from the object library with transformations specified by the control function. That is, the object, ground, or special effects processing channel takes the image stored in normal straight perspective and transforms it into a scene by changing the image, position, dimension, rotation, and distortion. Change based on the coordinates of . Image brightness is varied based on distance and object type.

The function of those parallel pipe twin processing channels is then to modify each object, ground and special effect used in a given scene as required.

Scene Construction The scene construction module takes individual images from each processing channel, separates the images from the background, and assembles a scene based on distance. In this way, closer objects crowd out more distant objects.

High frequency edges generated by constructing a scene from individual images can be smoothed by a Gaussian function. This action matches the edge and internal frequencies.

The scene construction module receives distance information from two object controllers. That distance is used to determine the force ≧force 1 that a particular object is in front of or behind other objects in the scene. If a particular object pixel is the closest occupied pixel in the scene, then that is the pixel that will be displayed. This is called "nearest" processing.

The scene composition function receives video input from each video channel and background level source λ determined by the FOv computer. In this function, the output can be a real-time video signal for special effects.

The digital scene composition function has the following sub-functions. 1
2) Light intensity value adjustment for brightness correction of the entire scene; 3) Smoothing to compensate for object-to-object boundaries and object-to-background boundaries.

Special effects A translucent special effect occurs after the sight occurs.

The special effects module adds special effects by varying the distance. Special effects such as smoke or dust control the transmittance of effects that occur before or after the image in the scene. The brightness value input controls the brightness/color of special effects such as black smoke or white clouds.

Databases Database hardware, like the database itself, can be separated into two subsystems, including object library hardware and gaming area hardware. The Object Library No. % -ware generates features in high fidelity and stores them on optical disks. Gaming region hardware is used to load locations such as objects, terrain, special effects, etc. The database nodeware then operates in non-real time to produce a high quality image on the controlled background. Those images are sent to an optical disk and stored there.

The library hardware consists of a disk controller, a disk drive, a signal conditioning module, and an optical disk. Either video (analog) discs or digital discs can be used for storing images.

Video discs are approximately 6-8 bits, or 64-256
Produces various gray tones. Digital discs can provide up to 12 bits of data. In all cases (with the exception of Senna images, where the image quality is very high), industrial 525 line non-contact video discs produce images of adequate fidelity.

The use of video discs is well known. The image is scanned by successive lines. It actually constitutes a column scan of the frame. However, as will be explained later, it is usually more efficient to start any of the warping processes described herein with the first pass sequence processing of the output of the video disc. Keeping this in mind, it is best to store the footage on a video disc with a 90 degree offset orientation so that the output of the disc is actually in the form of columns to facilitate first pass column processing. Therefore, it is desirable to do so.

Of course, if for some reason it is desirable to store frames on a video disk in their normal or vertical orientation, it can be easily done and buffered to suit the processing characteristics of the contortion sequences utilized. A front-end processing device can be provided for the processing channel to properly direct the data within.

Artists may think of other devices related to optical disks, but the CGSI concept has been found to work very well and is neither particularly difficult nor particularly expensive to implement. But me. Video discs have the following advantages:

a) High-density recording: approximately 54 per side of a 12-inch disc
000 frames.

b) Denture recording costs are relatively low.

C) Excellent constant-velocity callability: Certain improvements have made industrial disks faster than 60 cycle field times, or 16-(9
It appears that 50 to 100 frames plus or minus can be easily skipped every second. The actual skipping occurs during the blanking interval, so no image data is lost.

d) High data rate: produces data at video speeds; e) Long life and data security: the disk is non-contact and read-only, meaning no head collisions or operator errors can cause data loss. That's impossible.

f) Easy to replicate.

FIG. 2 shows a block diagram of the non-real-time database node 1-doquare. In this device, edges are restored, the background is separated from the object, brightness and color are corrected, realistic colors are generated, the object is positioned relative to the system's reference point, and non-real-time High fidelity CGI objects are generated and graphic data (light sources) are generated.

Database - Gaming Region Gaming regions include reference points that define the location of ground, objects, and special effects. Gaming regions can be set up in one of two ways: manual mode or automatic mode.

In manual mode, the operator searches the object library and selects which objects to place within the gaming area. Object files include bushes, trees, ground, mountains,
Individual objects such as roads, lakes, or groups of small objects on one file. To place a 2D object within the gaming region, the operator must
, y, and z. By the way, x, y
, z represent the horizontal axis, vertical axis, and distance axis, respectively. A second x, y, z reference point determines the height and position. Therefore, if the object is assumed to be standing upright at its true vertical position, then
and 2 are kept constant, and the Y reference varies by the height of the object. When an object tilts to one of those axes, at least one of the X(!:z axis reference points changes. The ground has four x,
The y and z reference points can be determined. Each of those points corresponds to one in each corner. This includes, for example, the sides of houses, lakes, roads, rivers, etc. To represent detail with great accuracy, three-dimensional multi-image objects can be stored in a series of images representing one degree increments in both azimuth and elevation.

They can be defined by three reference points representing the center attachment point, the center height, and the direction or pointing vector.

Automatic mode operates in much the same way as manual mode. Computers are devices that process and control objects. This will be explained later. Objects are placed according to type and density.

Database - Object Library The object library contains images. These images can be classified into three basic types: object, ground, and special-effects.

In the case of objects and ground, solid surface objects can be further classified into two-dimensional, three-dimensional-axis, three-dimensional biaxial, and light sources. Figures 4 and 5 show a method for recording objects and the ground on an optical disc with almost perfect high fidelity.

These figures will be explained in detail below.

Objects - Two Dimensions As mentioned above, most objects found in nature, such as rocks, trees, bushes, low trees, etc., can be represented in two dimensions with sufficient realism. A photo of the object is taken at the average aspect angle used in the desired simulation. As the altitude changes, the object is translated between reference points. Therefore, it becomes higher or lower depending on the angle of perspective. In the case of trees and bushes, the surface of the object is
As in the flight path, it remains perpendicular to the observer.

Experience has shown that this effect is not noticeable. The relationship of an object to other objects, its dimensions and changes in dimensions, as well as the rate at which those relationships change, give an indication of depth. For two-dimensional objects, a single photograph can be stored in a track on an optical disc and manipulated by twisting operations to obtain the appropriate rotation, dimensions and position.

Object - Assume that an object is tracked during a three-dimensional uniaxial flyover, the field of view changes by 90 degrees. In this case, the simulation may require an additional series of vertical axis photographs.

Objects - 3D Biaxial 3D/Biaxial objects or surfaces can be treated in three different ways. The first is to store a series of photographs representing small increments of 1 degree in both azimuth and elevation. This is a powerful presentation technique and works extremely well when objects have details that require high surface fidelity. helicopter,
Precise increments can be obtained by rotating a model of a large object, such as a tank or a house, on two very precise turntables and photographing the object at each setting. A second method of generating three-dimensional objects is by dividing the object into smaller surfaces, such as the house shown in FIG. A house such as the one shown in this figure can be seen in various perspectives in a series of views on a video display, but can also be constructed by a computer. A third method, applicable to large objects such as hills, is to take pictures of the object at a series of fixed heights and at relatively large intervals, such as 30 degree intervals, in an orientation completely surrounding the object. It's about taking pictures.

For example, you can use one altitude, typically 15 degrees, for most simulations, and 3 degrees around the object.
A series of photographs taken at 0 degree intervals has been found to be suitable.

Object-Light Source The light source is aligned from a series of points stored in memory,
A contortion algorithm contorts the surface from normal view to what is seen from the vehicle. This method works very well and has been used to make demonstration landing tapes.

Special Effects Regarding special effects, translucent objects or images create smoke, fog, shadows, smudges, and haze to add more realism to the scene. These objects can be stored as masks that determine their contour, shape, and transmittance. Masks determine the combination rate of objects and special effects. A second variable controls the brightness or color of the special effect. The mask determines the mixing ratio of special effects and variable control intervals fixed to the background. This technique can be used, for example, to generate dust that is kicked up after a moving tank. Twisting operations can also be added to distort special effects and the series of sequential frames used to create motion.

Therefore, a translucent object can be either stationary or moving. Special effect objects are defined in terms of transmission masks in the object library. This means that the data in the object library determines the percentage of background objects present and the percentage of special effects present by the following equations.

Pixel value (gray level) = ('z - mask) x (background (gray level) x (% special effect value - (gray value) The special effect value determines the gray shade of the special effect.

This is shown in FIG. Masks for static special effects can be easily drawn using a gray marker on a blank sheet of paper. In this way, many common or special clouds, clouds, smoke, haze, and fog can be easily drawn even by an inexperienced person. Special effects objects are typically processed as 2D objects. Multiple mask combinations can be stored in multiple rallies. Four types of special effects are realized as follows.

1. Moving Smoke The smoke mask, which defines the outline and transmittance, is generated by an engineer based on the photograph and the mathematical characteristics of the smoke. The top and bottom must be in the same place and must be the same width as shown at A in Photo XII. A series of frames, perhaps 480, are then generated. Each pixel can be incremented by one or more pixels in the Y axis as the frame is played back to create a continuous circular loop. This is shown in Photo XII, B. Next, shave off the top of the smoke in each frame (C in Photo XII) to simulate smoke dispersing into the atmosphere. The frames were recorded sequentially on a video disc as shown in Photo XII C, and 1
A contortion function within the special effects processor is used to inflate the top, crop the image to match wind speed, size the cloud based on distance, and position the cloud within the scene.

The first condition parameter sets the cloud color or brightness. The playback speed of the smoke determines how fast it flows.

2. Moving dust For example, 5-10 dust transmission masks can be made. Various masks (1-2, 1-3,...1
-10. A series of frames are generated by a series of linear interpolations between . . . 9-10). Those frames are recorded on a video disc. A contortion function in the special effects processing channel places the mask in the correct viewing position, determines its dimensions and position within the scene, and determines the color and brightness according to the initial setup conditions. This is shown in Photo XIII.

3. Yin Shadows are treated as translucent objects similar to dust and smoke.

A transmission mask for shadows is generated from images in the object library. One gray level that determines the transmission of shadows,
By setting all pixels in an object, we can create a transparency mask, a shadow. In the gaming area, four reference points of objects are projected onto the ground. The new point on the ground is the shadow reference point. The shadow, transmission mask is twisted to fit the scene based on the shadow reference point. This procedure is illustrated in FIG.

4. Shine and sparkle Shine and glitter are usually normal data on the surface.

However, in CGSI devices, surface normal data is not available unless the object is generated from CGI nodal data. To create shine and sparkle,
As shown in the following table shown in conjunction with FIG. 8, a fan-shaped mask is generated based on the areas of shine and sparkle generated by various solar angles. The sectors of the mask are gray levels. That is, when stored in the object library, the brightness value of sector 1 is 8, the brightness value of sector 2 is 16, and so on.

The solar angle table data is used to set up a survey table in the object processor. Assuming that the sun is within sector 2, input value 16 in the survey table sets the resulting brightness and twinkle values to predetermined levels. The remaining output values in the survey table are zero. As a result, a bright spot appears within sector 2. As the turret rotates or the sun moves, the fan shape changes. In this method, the movement of sparkles and sparkles is based on the movement of the sun and the vehicle.

Database - General Description The CGS I database is extremely flexible and consists of real objects and simulated special effects. The sense of realism achieved hereby depends on the operator's ability to configure a particular simulation. This database does not require a skilled programmer and is designed to naturally develop proficiency for a particular application. If a group operates in a certain area, ie, IR, visible, millimeter waves, or radar waves, the features from that sensor are loaded into the object file.

This figure simulates a sensor. This is because the image comes from the sensor. Furthermore, the parameters of the image or sensor can be modified during library creation or when setting brightness values during real-time processing. IR
An example is shown in FIG. In this figure, the selected emissivity can be varied to simulate changes to the sensor or many other parameters. If we are examining a particular area or type of terrain, we can load the features into an object file without the need for programmer assistance or tedious and expensive database changes. Contrary to the current state of the art in CGI equipment, the CGS I database is very well suited for rapid setup and quick training for a range of complex human factors situations. The flexibility of CGSI is demonstrated by the fact that the time of day, weather conditions, and most other parameters can be selected and easily implemented by experimenters without complete knowledge of the software. . Database - Description of 2D and 3D devices 2D natural objects The category "2D natural objects" includes trees, bushes, and small rocks. As mentioned above, one image drawn at an inclination of 15 degrees, for example, is sufficient for inclinations of 0 to 40 degrees and views at all azimuthal angles. Excellent visual effects
It is clear that the geometric relationship of any object to other natural objects in the scene and the change in dimensions. The static interior details of an object increase the realism of the scene by adding much more necessary detail. The average observer cannot perceive the fact that when a helicopter flies in a circle around a tree, the leaves do not change shape. However, leafless trees when used by helicopter pilots for hovering landmarks may look different and require multiple images in 3D.

2D Artificial Objects Artificial objects have a fixed orientation, such as to the front or side of the vehicle. However, in order to verify them,
We developed a simulation of a tank driving on a road with artificial or oriented objects (such as old trucks, cars, and tombstones) nearby. In this simulation, objects are always perpendicular to the viewer, but the 2D nature of those objects cannot be detected when viewing the tape. It appears that small to moderate sized objects can be presented at azimuth and elevation angles of plus or minus 15 degrees without appreciably reducing the quality of the image. However, a complete fly-around condition requires 3D processing.

2D Patching Technique Brush or small patches can be twisted and placed to form a very rough, wide continuous surface. For example, in a tank drive operation, the file of the cap can be made very large by assembling many small joints of the cap that are individually dimensioned and individually positioned to create a high degree of realism. Displayed well. In the helicopter portion of the demonstration videotape, the helicopter first takes off and flies over a patchwork field of grass. The use of patches represents a very powerful technique for adding highly detailed ground information. This technique can also be used on water surfaces, rocks, roads, railroad tracks, etc.

Furthermore, each patch can also perform movement.

That is, the grass or water is moved by the wind or the rotation of the rotor. This effect involves storing a series of dynamic frames on an optical disk, similar to the simulation of smoke movement.
It is simulated by feeding those frames to the ground processor.

This technique can also be used to represent objects at a certain distance. In the case of the helicopter, the cornfield shown in the "flyover" was constructed using several identical patches of cornfields. This same method can be used for dense groves, backgrounds, and any other very fine textured surfaces.

The entire area of the 2D surface textured feature can be twisted to create a textured surface. This concept can be extended to apply to long, thin objects such as runways, railroad tracks, roads, creeks, streams, etc. Linear twisting techniques can be used for approximately square objects J, but true perspective twisting should be used to realistically represent elongated "objects" to avoid image distortion.

The 2D surface need not be limited to the sides and can include light sources (point or line sources), rolling hills, sand dunes, ponds or small lakes.

The urban area simulated using CGSI technology can consist of the ground, streets, sidewalks, and building fronts.

3D multiple surfaces allow most man-made objects to be subdivided into many surfaces, such as the side of a building or a truck. 2 on each side
A 3D object can be created from a small database of individual sides by treating it as a D-surface and allowing the computer to construct the visible sides of the object within the scene. For example, consider a house that has two side, two end, and two roof sections. Each section is determined by four (x, y, z) reference points. Therefore, the number of coordinates is (4 corners) times (3 reference points) times (6 sides) = 72. These coordinate numbers are necessary to locate all the details of the house within the gameminder area.

The many views of objects required by CGS I technology are
Estimates can be generated from dimensional data or obtained from digitized models from actual photographs of actual parts.

3D multi-view complex high-fidelity object images lose fidelity when rendered into flat or curved surfaces. Two examples are a leafless tree with branches and a tank with many irregular surfaces. These objects can occupy more than a third of the screen area, but can be stored as a long series of 2D views with azimuth and elevation increments as small as 1 degree. The width of this multiple view requires approximately (90x360) 32400. This number of frames accounts for about 60% of a video disc holding 54000 frames per object.

The most popular airways shown in Figure 10 are those that encircle an object while changing elevation.

If on a video disc the frames were indexed with angular values from 0 to 9 degrees of elevation in 1 degree increments (for each 1 degree incremental change in azimuth), then if the flight path was almost flat, However, the disk must jump about 100 frames to reach the next frame of interest. During the vertical retrace of each TV field, the disk must be able to jump about plus or minus 100 frames. This makes the circumference of the object 6
Various elevation angles every second (360 degrees/6o field/second)
You will be able to fly. This is a realistic limit.

This is because even a 360 degree turn in exactly 6 seconds can make a pilot dizzy.

Other ideas have been proposed, including linear and non-linear interpolation methods for 30 multi-view objects, but to date none have been successful in achieving the same level of detail as the present invention.

Special Effects Special effects are processed similarly to 2D objects and surfaces, except using transmission masks. The mask can be dynamic (changes each frame, similar to those for flakes, smoke, etc.). You can also control the color and brightness of the material. All those special effects techniques are demonstrated in CGS I video output mode.

According to the concept of occultation CGSI, objects may have holes or windows. That is, for example, the next object or background can be seen with determined realism through a gap in a branch. Features or parts of them that are closer to the observation point always obscure features or parts of them that are further away. As previously explained, the scene construction module uses the distance data to select each vixel in the final scene. Three examples are described to illustrate the occultation performance using distance data. Examples thereof are shown in FIGS. 11-14.

In the first example shown in FIG. 11, the shading of the three trees can be based on distance data to a reference point in the gaming area. In this example, tree T0 shadows trees T2 and T8, and tree T shadows tree T8.

In the second example, an elongated vehicle is moving around a tree. As shown in FIG. 12, using a single range point for the vehicle results in inaccurate occultation. In Figures 12A and 12B, the vehicle is in front of the tree. However, at B, Rv is greater than k, so the vehicle is placed behind the tree. This method is not sufficient. In Figure 13, two distance data points are maintained for the vehicle, a maximum and a minimum. In this case, the distance of the vehicle can be determined by using the distance vector of the vehicle, which is close in magnitude to the distance vector of the tree. FIGS. 13A-13D show an example of a well-placed vehicle moving around a tree. A third example deals with a problem between two vehicles, as shown in FIG. In this example, the distance can also be correctly selected by using the closest vehicle vector. This method can address any number of objects.

2D/3D Conclusion The above discussion provides an overview of 2D and 3D methods for placing objects and special effects in a scene, and techniques for implementing them. With these techniques it is possible to simulate most objects and conditions encountered in the real world. Appropriate 2D/
Applications of 3D technology will be apparent to those skilled in the art.

The hardware and software for generating the location of a vehicle simulation trainee's vehicle or other scene recognition device relative to the gaming area and the location of objects moving within the gaming area, as such, constitutes a CGS I device of the present invention. k in part
stomach. However, the x, y, z that indicates the instantaneous position of those vehicles
The roll, pitch, and yaw signals of the CGS I device are detected by the CGS I device, and in response to those signals, the video display generates a scene of the simulated gaming area.

Communications Subsystem The hardware for interfacing CGSI devices to vehicle simulators is unique for each application. This seems to be the case. Each user has a determined vehicle simulation computer and the CGSI device can have a standard sword. The simulation
Converts the digital output signal from the computer to cG
It is the interface hardware that matches the input to the sIFov computer. This hardware ranges from simple 7" cables to complex interfaces including buffers and microprocessors.

Field of View and Coordinate Transformation Calculations Figure 15 shows a diagram of the FOV function for the gaming area. This is the origin 20 of the gaming area coordinate system, and the observer's eye is at an arbitrary point 2°1 within the gaming area. The screen or CRT placed immediately in front of the observer, i.e. the v11 exerciser, has a contour 22 and the origin of the coordinate system of the screen is a point 23. The four corners of the patched part of the terrain are projected by a CRT. The patch is the windshield of the aircraft, and the contour 24 represents the projection of the terrain patch on the screen. It can be viewed by the operator.

Two types of vehicle simulation calculations are provided to the FOV processor: (1) position vectors that define the changing position of the aircraft relative to the origin of the terrain coordinate system; and (2) position vectors that define the changing attitude of the aircraft relative to the origin of the terrain coordinate system. Rotation data (yaw, pitch, roll). The formula representing this data is shown in the table below.

The eye coordinates expressed in world system coordinates tZ = X θ = Yoty = y φ = Pitch tz - Z ψ = Roll + ■ include the following Vll V12 Vll! V21 V22V28 sinψ Vllll VB2 vaa cosψV41V&
2 V48 Table 2 The vehicle data in the form of the equation shown in FIG. 16 is processed according to known principles to determine the distance of any object or ground, such as a patch of terrain, from the screen 22; The screen coordinates of the four corner points or the vertices of the projection of an object or surface onto the screen can be generated. In fact, the performance of the prior art is that a patchwork of terrain or any object or surface has convex polygons of any shape and dimensions, and the coordinates of the vertices of those polygons are It is possible to calculate .

However, there is no need in the present invention to take advantage of this advanced capability of the prior art for storing objects, surfaces, and special effects in a database.

As explained above and shown in FIG. 16, photographs of objects and ground are stored on a video disk, and the recall of these stores is controlled by an index controlled by a FOv processor.

Each object or surface has its own input image within the frame. The frame size of the video is 512 pixels or pixels (1 scan frame).
It has two scanning lines.

The database has a table of all objects, all surfaces, and all special effects (individually indicated overall by the abbreviation 08SE) within the gaming area. The locations of those objects, etc. within the gaming area are indicated by gaming area coordinates. The database also contains information about the height of objects within it. The FOV software allows 08SE in the field of view and 08SE from the video screen.
The distance of each SE can be determined in real time.

In FIG. 17, the shape of the frame 25 containing the input image of the object depends on the true shape of the object, and the FOV processor uses the table to determine the corresponding screen coordinates 1-4 of the video screen 26. The height is used in the conversion equation 2. Intermediate image 27 is an important part of the linear twisting algorithm. The "algorithm facilitates the mapping of video from input frames 25 to screen frames 26, which will be discussed in more detail below.

In summary, for the FOv function, by processing data from vehicle simulation calculations, (1) the 08SEs within the field of view are determined, and (2) the distance of each 08SE from the observer's location is determined. , (3) The screen coordinates of the four vertices for the closed space into which each 08SE input image is mapped are determined. The above data for each 08SE is provided to the 08SE processing channel, as explained below.

Object/Ground/Special Effects (O8SE) Channel The processing channel or ossa channel processes object data, ground data, and special effects data from the database library. As mentioned above, the same channel hardware is appropriate for all three functions.

The 08SE channel is important and essential for CGSI equipment. A possible hardware implementation of the 08SE channel is shown in FIG.

Several functions are performed on the library data by the 08SE channel to obtain the correct brightness, color, image, dimensions, position, rotation and view, as described below.

a) A high speed (approximately 100 nanoseconds/sample) analog-to-digital converter 30 converts the image of the object into digital format. Typically, digital formats have 480 effective scan lines (525 total) with 512 pixels each. 1
Each pixel includes 8 bits (256 gradations).

b) High speed memory card 32 receives X@ or Y axis digital data. The loading axis and direction depend on the image rotation. Data is loaded to minimize pixel compression during processing. For example, if an image is
Instead of rotating 0 degrees (rotating 60 degrees would result in some image loss), the data is loaded on the vertical axis (90 degrees) and rotated by 30 degrees.

Memory card 32 also holds images of objects for processing when the optical disk controller is selecting a new track (image). This card can be omitted if the object is stored on the disk in 90 degree increments or if the rotation is less than +45 degrees.

C) The research tool 34 modifies the image brightness values to obtain distance and contrast effects. This operation requires a delay of several pixels.

d> Twisting card 36 transforms the Y-axis image line by line. The starting point (offset) and scaling factor move, compress or stretch the pixels of each line. This operation delays the flow of pixels by one line.

e) A second identical high speed read/write X and Y axis memory card receives and stores the converted Y data for the odd and even fields to form a frame. . After the Y-axis field is loaded into the Y-axis, the X-axis data is read out by line and by the even and odd fields. This buffer operation requires one video frame.

f) A second twisting card 40, identical to card 36, processes the X-axis data for line translation, stretching or compression.

The brightness controller of the brightness control survey table includes a memory controller and an object survey table 52. During the active part of the display time, the survey table, , , (L
! The input video is used to address the JT, ) and the data output of the LUT is sent to the contortion function configured on the card 36, 38, 40 for further processing. In this procedure, the input brightness value is I,
It is effectively mapped to the output luminance value via the data stored in the TJT. During the vertical blanking period, the memory controller 5o can take addressing control of the LUT 52 (when so directed by the object controller) and fill the LUT with a new set of values to form a new object. or modify the appearance of previously selected objects.

Brightness control is divided into two functions: a function that performs brightness corrections related to a particular object, performed by the card 34, and a function that performs brightness corrections related to the overall scene, as will be explained later. .

The memory controller 50 can be implemented by a one-chip microprocessor, and the LUT 52 is a RAM with multiplexing and control circuitry to provide access from the video data and the data in the memory controller 5o.
It can be configured as

Linear Thread Bending Technique One linear twist technique related to the present invention is the card 36
．． 38.40 and includes a procedure for performing a spatial transformation on a digital image represented as a matrix of brightness values. Given a rectangular input image accessed from a database on an optical disk, this technique draws the four corner points of the output image in a straight line on the video screen. This is done in two orthogonal passes, as shown in FIG. Each pass linearly interpolates each input line to a different size and positions it in the output image. The dimensions and position parameters for each interpolation are determined from the input image corner coordinates and the exit corner coordinates. This interpolation consumes successive input pixels to produce successive output pixels. The two passes interact to perform dimension transformations, translation transformations, rotation transformations, plus non-standard mapping.

This process is independent of the FOVO formula that calculated the four output corners on the video screen. Once the four output corners are set, they are computationally invariant to all transformations. Since it operates on streams and lines directed to successive sequences of pixel values, it is ideally suited for real-time hardware implementation.

Each pass simply dimensions (grows or shrinks) the input line,
It simply positions (offsets) its input line within the output video image. These two operations are performed by continuous interpolation over discrete fields of pixels. This allows for continuous sizing of lines and sub-vixel positioning of output lines and columns. This completely eliminates naming of diagonal edges. At the heart of this technique is a method that allows the device to sequentially size and adjust the position of individual input lines relative to individual output grids.

Interpolation FIG. 21 is a flowchart for the continuous interpolation operation for the process shown in FIG. In Figure 21, symbol 5
IZFAC is the "size factor" applied to the input image line, and lN5FAC is the inverse of 5IZFAC, with the additional meaning of indicating which part of the input pixel is sought to form the output pixel. . INS
EG is the portion of the current input pixel available to contribute to the corresponding output pixel, j5.0UTSEG
refers to the portion of the output pixel that still needs to be completed.

According to the above definition, the p1 process has lN5EG and 0tJTS
It begins with a comparison of values with EG. OUT S EG is IN
When it is less than 5EG, it means there are enough input pixels available to complete the output pixels.

Conversely, when lN5EG is less than 0UTSEG, it means that there are not enough input pixels left available to complete the output pixel.

Therefore, the current input pixel is always used and new pixels must be drawn to complete the output pixel. Only under those two conditions will an input pixel be used forever without completing an output pixel, or an output pixel be completed without using an input pixel.

Assuming the output pixel is left to be completed, the current pixel value is multiplied by 0UTSEG and the product is added to the accumulator. lN5EG is 0U to indicate the usage of that portion of the input pixel.
is decreased by the value of TSEG, then 0UTSEG
is initialized by the IN5FAC to indicate that a complete output pixel is reserved for filling. The contents of the accumulator are scaled by 5IZFAC and the result is the value of the next output pixel. The process then returns to compare the new values of IN5EC and OU'I'SEG.

Assuming there are input pixels left to be used,
The current pixel value is multiplied by lN5EG and the product is added to the accumulator. 0UTSEG is lN5EG
to indicate that the portion of the output pixel has been filled. Then IN5EG is reinitialized to 1.0 and the next input pixel is retrieved. The process then returns to compare the new values of IN5EG and 0UTSEG.

The core of this process is the interaction between IN5EG and 0UTSEG to scale input pixels to output pixels. The effect is one of scaling the pixels and moving them from one individual grid to another by a continuous interpolation process. Of course, the success of this continuous scaling process depends on the small accuracy of IN5EG and OUT S EG. A sufficiently high precision has the effect of a completely smooth pixel movement between the two individual grids.

Subpixel fading is performed by initializing 0UTSEG to a percentage of INSFAG to create a partial first output pixel. When the input pixels are used up, the last output pixel may not be completed, resulting in the final output pixel being partial. This allows the output pixels to be placed consecutively in relation to the individual pixel grids, and avoids having separate names for the edges.

By sequentially sizing and phasing the individual input lines with respect to the individual output grids, the quadrilateral twisting can be done with each column of the first pass and the second The dimension parameters, phase parameters, and output location parameters for each line of the path are determined.

The first pass (Figure 16) reads the input image and writes intermediate images vertically from left to right. The object moves all pixels along their correct vertical axis. This is done by mapping the first column between Y1 and Y2 and linearly interpolating to all other columns such that the last column starts at Y2 and ends at Y3. .

The second pass reads lines from the intermediate image and writes lines to the output. Since all pixels are now in their correct rows, rows can be processed independently in any order, such as interlacing 2 to 1 fields. The purpose of the second pass is to move all pixels to their correct horizontal axis orientation.

The second path mapping must be considered in three processing regions, as indicated by the dashed lines in the intermediate image. There is a different output location delta for each region, and only during intermediate region processing the size factor is updated by that delta. With the correct initial values and deltas, each corner will be mapped to its output X coordinate and the rest of the image will follow the appropriate relationship.

The initial output location is the intersection of the 1-4 edge and the horizontal line passing through the top corner. The location delta for the top region is the edge slope of 1-4 if corner 1 is the topmost and corner 2
If is the topmost, then the slope of the edge is 2-3. The location delta for the intermediate region is the edge slope of 1-4.

The location delta for the lower region is the edge slope of 1-4 if corner 4 is the bottommost, and if corner 3 is the bottommost
2-3 edge slope. The first size factor is the length of the horizontal line segment from the second highest corner to the horizontally opposite edge. The dimension delta is the value subtracted from the similar value of the third-highest corner, divided by the vertical distance between the second-highest corner and the third-highest corner.

In this way, corners 1, 2, 3.4 of the input image are mapped to corners 1, 2, 3.4 of the output image. This mapping may require preorienting the input image by 90 degrees, 180 degrees, or 270 degrees prior to the mapping operations described herein. This orientation is determined by calculating the areas of four possible intermediate images and selecting the orientation that maximizes the area. After predetermining the orientation, the corners of the input image and the corners of the output image are relabeled such that corner 1 is the top left corner.

Then, the twist function implements the process using row-by-column passes and line-by-line passes. FIG. 19 is a top level block diagram for the twisting function with three subfunctions defined. The same object processor 60.62 located on the card 36.40 is used. Each processor can perform one-dimensional twisting. The X object processor 60 performs column-by-column twisting and
Object processor 62 performs a line-by-line twist on row p0 to allow X object processor 62 to access row data rather than column-directed data in the serial data stream that had been used up to that point. buffer 64
is an intermediate image (twisted in the Y direction instead of the X direction)
Used to store.

The magnitude and offset parameters for the Y and X passes essential to the implementation of this algorithm are sent at frame rate from the appropriate object controller to the Y-axis processor 60 and the X-axis 7' processor 62. Line-by-line (-!, column-by-column) calculations of magnitude and offset must be done in the axis processor itself.

Realization of linear twisting technology As shown in FIG. 20, the input memory 70 and the first pass
Processor 60, intermediate memory 64, and second pass processor 62 are used to implement a two-pass technique in a network architecture. The intermediate memory is the first pass processor 6
While 2 writes its results into one buffer,
The second node processor reads its inputs from another node (double, so that it can be read from another node). Both process stages can be of the same 71-ware design.

- The subsystem is provided with three external communication points. The input image memory 70 is loaded from the image database and is a 10 MHz, 8 bit I10 port. The host must send sub-image coordinates and output corner coordinates to controller 72. This is a high speed PS232 or shared DMA port. This third port is the output image presented as a 10 MHz % 8 bit, synchronized, 2-1 interlaced stream of pixels.

The host must configure the input image as shown within frame 25 of FIG. 16 and provide image corner points 1-4 as shown within screen frame 26. This subsystem then performs the directed transformations on the supplied images until the host changes one or the other.

Control and processing within the subsystem is naturally divided into three hierarchical levels: frame level, film level, and pixel level.

Frame processing and line processing have lower bandwidth than pixel processing and can be performed on available 16-bit microprocessors. Pixel processing requires specially designed hardware because the data rates are very high.

Frame level control includes setting up memory, communicating with the host and starting the line processor. Several control tasks are performed at the 7v-me level. The orientation of the input stage access is determined by finding the orientation that yields the largest intermediate image. This is illustrated by frame 27 in FIG. There are four possible orientations and output positions, and the size factors and associated deltas must be calculated and sent to the line processor of FIG. 20. For the second pass, the calculations are somewhat more complex than those for the first pass, and in addition, some screen clipping calculations must be performed.

33 ms of time is available for performing frame-directed computations, and by performing double-integer arithmetic, those tasks require only about 15 ms in M2SO4, for example. It is. Therefore, one microcomputer has sufficient processing power to perform these calculations and control the subsystems.

Line level processing increases the output position and size factor, performs clipping calculations on the output position, sets addresses in input and output memory, and signals the pixel processor to begin processing. The time available for this process is 63 microseconds. It is estimated that the conflict will take at least 50 microseconds. Since this is close to the limit, two microcomputers may be required to ensure line calculation performance for each path.

A separate line processor is required for each pass, but only one frame processor is required. Therefore, a total of five microprocessors are required. Each process requires very little data storage and the required program is short.

Therefore, a very small amount of memory is required, and a total of five microprocessors can be mounted on one board.

FIG. 22 shows a typical interpolation processor for performing an interpolation method that consumes input pixels to generate output pixels. This processor implements a two-state interpolation loop technique that can be easily implemented in hardware with a pipeline structure.

Output pixels could easily be generated at a maximum rate of one every 100 nanoseconds. Since this algorithm requires two cycles to generate an output pixel, the pipeline must be able to operate in 50 nanosecond cycles. This process requires minimal hardware and is compatible with high speed pipelines.

A recursive combination 75E exists for a single pipeline stage.

This is a comparison between IN5EG and 0UTSEG. Using 1 to scale the input pixel (INSEG) and subtracting 1 from INSEG based on the comparison results in a new value for the 1 that is then compared again. The coefficients selected for scaling are input to the next stage register under clock control and held there. However, in one stage, comparison and subtraction are always performed during one cycle.

This is shown in FIG.

If the size of the first stage motion image is to be reduced, 0UTSEG is smaller than IN5EG, and the output pixel is completed. OUT S EG is stored in the coefficient register of the next stage,
Derived from lN5EG. lN5E) The new value of G is stored in its register. OUT S EG is lN
5EG 75'' are initialized again and comparison 75S is set for the next d cycle.

On the contrary, lN5EG is more sensitive to J-
When the input pixel is completed, the input pixel is completed. lN5EG is placed in the coefficient register of the next stage and subtracted from 0UTSEG.

The new value of 0UTSEG is stored in that stage's register, lN5EG is reinitialized to 1.0, and the NO<iB line stage is set for the next cycle. The remaining stages perform similar operations.

Second Stage The second stage multiplies the input pixel value by a selected scaling factor. When the input vixel is exhausted, the next input vixel is input into the vixel register. The result of the multiplication is provided to the next stage of the pipeline, the third stage.

3rd stage The 3rd stage accumulates the scaled a-square of the input pixel -1-
Assuming that all 4° input pixels have been used, the process ends. When the output pixel iE is completed, the accumulated value is given to the next stage and the accumulator is cleared to zero.

Fourth stage The fourth stage shifts the value given from the third stage and normalizes the position of /J% several points in order to input it to the output standardization multiplier. This normalization will be explained later.

Fifth Stage In the fifth stage, the accumulated value is multiplied by 5IZFAC. This is the opposite of IN5FAC. This operation yields the value for the next output pixel. The value is applied to memory and stored in its next output pixel location.

Arithmetic precision interpolation is sensitive to the precision of both the output luminance values and the spatial orientation of those values. Computer simulations have shown that a value of 8 fractional bits is sufficient to achieve very high output fidelity over a wide range of conversions.

Since the value of INIG never exceeds 1.0, 9 bits are sufficient for display purposes. 0UTSEG can be represented by 16 bits, of which 8 bits are below the decimal point. It is easy to see that the smaller of the two values is always selected as the conversion factor for the next stage.

This means that there will only be a conversion factor of 1.0 or an 8-bit fraction thereof.

A coefficient of 1.0 can be detected and treated as a special case that bypasses the scale multiplication and feeds the pixel value directly to the accumulator. This leaves only the scale factor with 8 fractional bits.

For scaling, an 8x8 multiplier is sufficient, so the product is a 16-bit value with 8 fractional bits.

Accumulators accumulate their values. There are also two possibilities. If lN5FAC is less than or equal to 1.0, the associated pixel bits will fall into the fractional part of the product, resulting in only one accumulation.

On the other hand, if lN5FAC is greater than 1.0, there may be several accumulations because the associated pixel bits go into the higher order bits. Therefore, the accumulator must be 24 bits wide and include 8 fractional bits.

The accumulator value is progressively scaled to the final output value by the 5IZFAC. Only 8 bits of this final scaling are required. The 8 bits to the left of the decimal point are the desired bits. The possible values that can be accumulated are lN5
It is directly related to the values of FAC and 5IZFAC. Therefore, the value of 5IzFAC is directly related to the range of possible values in the accumulator, and it is possible to determine the associated bits in 5IZFAC and the accumulator that produce the desired 8 bits.

5IZFAC can be normalized so that its high order bit is 1. If the value has its eight or more significant bits truncated and the resulting decimal point is noted, the resulting value may contain, for example, six decimal bits. Based on the value of S I ZFAC, the relevant bits are taken out of the accumulator with 8 fractional bits. The retrieved value has enough bits to bring the total number of fractional bits between it and 5IZFAC to 8. When two numbers are multiplied together, the 16-bit product has eight fractional bits, the desired bit to the left of the decimal point in the upper 8-bit output pixel value.

A dual-axis high speed memory memory functions for the pixel processor by sending and receiving streams of successive pixels in designated rows or columns. The line microcomputer loads the row and column addresses of the image into memory address columns.

The line microcomputer specifies whether the address should be incremented as a row or column and whether a read or write operation should be performed. The memory responds to requests for the next pixel from the pixel processor. For each new line, memory is loaded by the line microprocessor to the new row or column address.

The memory must respond to requests at a maximum rate of once every 100 nanoseconds. This speed is achieved through a technique that involves consolidating the memory into four banks of 8 bits each. In this way, one memory cycle has 4
pixels can be supplied or stored. Pixels can then be moved four times faster than the base memory speed allows. As shown in Figure 24, by fast recalling an image in a row and column, and mapping that image into memory in a spiral manner, four consecutive pixels in any row or column are stored in separate memory banks. This is achieved by doing more.

Image address-memory address mapping No. 24-2
FIG. 6 shows how image address-memory address mapping is performed. Row and column image addresses are stored in up/down counters and internally mapped to memory addresses. Matupingyosi,
This involves increasing the row address for a column-directed call, or the column address for a row-directed call.

Determining the lower two bits of the 16-bit memory address,
Controlling the rotation of the address orientation of input and output registers for each memory bank is a more difficult aspect of address mapping.

The upper 7 bits of the 16-bit memory address include the upper 7 bits of the 9-pit column address. The lower two bits of the memory address are responsible for the row 77 column flag. For row-directed recall mode, the lower two of each bank
The bits are the same, the lower two bits of the row address. For column-directed calls, the lower two address bits are rotated according to the lower two bits of the column address, which are different for each memory bank.

Lower 2 memory bits Lower 2 column bits Bank A Bar/B Bank C
Bank D00 00 01 1
0 110'l 11 00
Q 1010 10
11 00 0111 01
The control of the 10 11 00 car output register is the same in both row and column access modes and is a function of the lower two bits of both row and column. Register address assignments rotate depending on the row or column being called.

Considering only the lower two bits, the zero pixel for the zero column is in bank C, and so on. The same applies to rows. This rotation register address has 0 in bank A,
This can be achieved by assigning 1 or 0 to each bank, such as 1 to bank B, and adding modulo 4 to the lower two bits of the column and row addresses.

For each request from the processor, the appropriate row or column address is incremented. The pixels are multiplexed from the output register or multiplexed before entering the input register. Lower 2 of increasing addresses
When the bit changes from 11 to OO, a memory cycle is initiated to store the 4 manual pixels or read the next 4 pixels.

Large/Small Channel Processor CGS I To construct the scene, each object, ground, special effects (ossg) image is passed through the processing channel as previously described. The dimensions of each 08SE range from pixels to the entire screen. In most scenes,
Small objects (for example, 1/16th of the screen area or less) and large objects (for example, 1/16th of the screen area or more)
There are many objects. The processing channels are configured to operate on an entire frame. They are called large channel processors. Therefore, only one of the frames
Using parts as small as /16 is completely inefficient. In this case, there are three options to choose from. That is, using large frame processors in parallel with low efficiency, creating a dedicated small 08SE processor, or processing multiple, say 16, ossgs with a large channel processor. Of course, the last choice is the least practical.

Multiple small 08SEs can be processed by a large channel processor in parallel or series, as shown in FIG. In case of parallel, 16 small 08
The SE is loaded into one memory plane into the load plane. , ossg are located in respective memory cells of the memory plane.

The X and Y screen addresses of the cells and the 08SE addresses within those cells determine the position of the image in the eyes of the scene composition module. In the serial method, all the small images enter the processor first in Y and then in X to be processed. This serial method uses 16 small input and output memory planes.

As explained earlier, one inherent problem with CG engineering image generation technology is that when transitioning from an object to a background, the edges lack smoothness, resulting in a cartoon-like appearance. That's true. One notable advantage of the continuous interpolator of the present invention is the elimination of artifacts in the edge detail of pictorial images. For example, a painterly image obtained by photography reproduces objects realistically, and the transitions between each object and the background, which have many boundaries, are not abrupt but light transitions. The interpolation characteristics of the present invention: - Faithfully reproduce the subtle luminance changes that occur between the object and the background. This is in sharp contrast to the prior art, where adjacent pixels have widely different brightness levels, giving the appearance of sharp edges and creating a cartoon-like effect as described above.

Perspective Bending TechniquesFor straight bending, the advantage of interpolation techniques is that most of the highly variable calculations can be done at the frame level, where most of the time is available.

Line calculations are very simple, and pixel interpolation can be done with just a few parameters. Although the linear interpolation method can obtain an accurate contour of the perspective transformation pool, such mapping is not accurate since the mapping of internal data is non-linear.

This non-linearity is manifested by continuous interpolation in the form of continuously varying size factors of each new output pixel pool. That is, in the linear transformation, the value of lN5FAC is constant over the entire column or line, whereas
In perspective transformation, the value of lN5FAC may be different for each output pixel.

It is desirable that the function for changing the value of IN5FAC or for changing the size factor be simple enough to compute so that it can be incorporated into low cost real-time hardware. This can be done with a two-bus operation. This considers vertical and horizontal paths in terms of vertical and horizontal planes instead of rows and columns. The first pass uses a series of projections of each object column onto a vertical plane, and the second pass uses a series of intersections of the object with a plane projected from the origin through each line on the screen. .

The calculations for the plane are shown in FIG.

The observer's eye at point 0 is the origin of a coordinate system. Straight line AB represents the screen. Straight line CD is a line representing an object in the space behind the screen.

Lines emanating from the origin indicate line-of-sight rays intersecting screen AB in 1-pixel increments. Those rays also intersect straight line CD. The point where a ray intersects the object line is mapped to the point on the screen where the same ray intersects the screen.

Therefore, the portion of the complex line between the intersections of any two rays is the portion of the input image that is mapped to pixels on the screen between the intersections of those two rays and the screen. The length of the object line segment between any two rays is the number of input pixels, ie, the value of lN5FAC. Its input pixels contribute to the corresponding output pixels.

l for each output pixel for the interpolating power processor
To obtain the value of N5FAC, K needs to trace the rays from the beginning to the end of the object line, solve the intersections of each ray, and know the distance between the intersections.

The intersection equation is very simple and contains constant terms in most parts. The ray passes through the origin and is characterized by the distance to the screen and the slope of the ray, which depends on the coordinates of the screen. The screen coordinates are the only value in the equation that changes as the ray is traced.

The object line equation is characterized by its end points.

MOJB is defined as the slope of a straight line.

The intersection equation is solved for the two coordinates of the intersection of two straight lines.

By subtracting the two coordinates of the previous intersection, the horizontal distance between the two intersection points is obtained.

From FIG. 28, the object line CD forms a right angle in the coordinates, and the segments at each intersection point also form a right angle. The ratio of the length of straight line CD to the horizontal distance between points C and D is equal to the ratio of the length between each intersection to the horizontal distance between those points. This constant ratio RATIO can be determined directly from the end points of the straight line CD and the length of the input line, and can be applied directly to the length of the line segment and the horizontal distance between the intersection points to determine the following value of lN5FAC: .

This calculation consists in applying the varying screen coordinates via equations. As shown in FIG. 30, this can easily be implemented as a pipeline configuration in hardware. For each new pixel, the pipeline
Provide the new value of AC to the interpolation processor.

The exact object line CD for each column and each row can be characterized by the projection and intersection of the objects in three spaces onto the plane.

The first path shown in FIG. 31 is in the vertical plane YZ and
The projection of a row of objects onto a plane is used.

This can be done by determining the YZ coordinates of the top and bottom of each pixel column in the image of the object in three-dimensional space. The YZ coordinates of those two points in each relevant column are the end points of the object line in the yz plane traced by the ray in the interpolation method, as represented in FIG.

The second pass (Figure 29.32) uses a plane that is projected onto the screen through each line leading from the origin.

The intersection of that plane and the object re-defines the straight line of the ray-tracing pass. This straight line is defined by the end points of the straight line in the plane and is found by finding the intersection of those end points with the plane of two three-dimensional straight lines formed by the edges of objects in three spaces. .

A two-pass method that utilizes rays tracing along projected object lines to perform continuous interpolation achieves a mathematically accurate, high-fidelity perspective transformation of a planar image onto a screen. It is something. This is shown in Figure 35 for both the straight and perspective cases.

The processing of images in strabismus was explained in the technical section. In order to better understand the perspective distortion of an image, an explanation is given below.

It was previously explained that any frame projected orthogonally or obliquely onto the screen is stored in memory in the same manner. Frame data is 512
A pixel array of 512 rows and columns, each associated with an intensity value.

In the case of strabismus, the FOv program calculates the coordinates of the four corners of the frame in two spaces. The bispace is the coordinate system with the receptors of the scene recognition device.

The receptor is an observer, and is treated as the origin of the coordinate system. An example of frame 100 in two spaces is No. 31.
This is shown in Figure 32. 28th and 29th.

Referring to Figures 31 and 32, the origin, ie, the screen AB in front of the observer's eye 0, is an xy plane that is a predetermined distance away from the origin.

As shown in FIG. 31, the 1000512 columns of frames are sequentially projected onto the straight line CD in the YZ plane.

The YZ 'l' plane has no direction with respect to the X axis. In each case, the straight line CD (object line) is one side of a right triangle, of which each column of the frame forms the hypotenuse. The interesting property of each object line CD is the end point of that line in the Yz plane. The coordinates of the end points are calculated and stored in memory.

A vertical path operation shown in FIG. 29 is applied to each of the 512 object lines CD. The operation is performed in the YZ plane. A line of sight 101 exits from the origin, that is, the eye 6, and intersects with the screen AH. The distance between a pair of observer's eyes is 1 pixel. The line of sight 101 also intersects with the object line CD as shown.

The object line CD is obtained from a frame sequence with 512 pixels and is assumed to have a corresponding length converted to 512 units for calculation purposes.

Line segment 1 of straight line CD formed by intersecting lines of sight 101
02 is an indefinite length, the length is 5 for straight line CD
Based on the total length of 12 units, the following Table 3 shows (Table 3) Y = MRAY 2 Line equation of ray Y-YD = MOBJ (Z-ZD) Line equation of object (constant) INSFAC = (Z- PREVZ),,,
RATIO Once the length of the line segment 102 that maps to the next screen pixel is known, the assumed total length of straight line CD 512 units is proportionally distributed to the individual line segments. At this stage, the pixel ratio between each line segment 102 and the corresponding line segment on screen AB is obtained.

This ratio is the lN5FAC value for each line segment 102.

Since the lengths of line segments 102 vary from one another, the value of lN5FAC for each line segment varies correspondingly.

By using the stored pixel intensity values of each frame row, we can calculate the corresponding line segment 105 in screen AB for the intermediate image using the sizing and interpolation methods described above in connection with linear twisting. obtain.

By repeating the above operation for each of the 512 object lines CD, an intermediate perspective view is obtained. In this intermediate perspective view, the 512 vertical lines corresponding to the rows of 10 frames constitute a properly sized and positioned vertical component of the final image. The intermediate image does not have a horizontal orientation.

In the second stage, shown in FIG. 32, a plane 103 extending from the origin and across screen AB in rows 104 intersects frame 100. The length of row 104 is equal to the size of one pixel. Both ends PI of the straight line where the half surface 103 and the frame 100 intersect.

The line segment 111/q of Q' is somewhat similar to the straight line CD of the vertical bus.

Because the columns of the intermediate image are separated by one pixel after the first pass of processing, the intermediate image has a particular width in memory. Its width is somewhat arbitrary. Therefore, the in-width field is not important except that it is used in the second pass calculation to obtain the final image.

Line segment FIq can be considered as a second path structure line. The reason is that each structure is projected onto the XZ plane that includes the origin. This half plane is the Xz plane shown in FIG. In FIG. 29, line F/Q is shown as line PQ.

The width of the intermediate image corresponding to each structure line Flq is equal to the length of the second bus object line PQ. Screen ABO line 104 is a projection from line Flct and is the final output line.

A straight line passing through the intermediate image at the same level as line yq is mapped to a corresponding straight line 104 on the screen.

4 Now, referring to Figure 29, set the origin and screen AB.
' and (separated by one pixel), a vertical plane 106 passing through one of the second pass object lines PQ connects the line PQ to a series of line segments XI, X2 . Divide into X3... Their dimensions are inherently different.

The length of straight line PQ corresponds to the width of the intermediate image at the vertical level of straight line 104. The pixel columns of the intermediate image are equally spaced. The reason for this is that the horizontal spacing of the columns of the input frame 100 was disturbed by the first pass operation.

Line segments XI, X2 . etc. represent a number of intermediate image pixels corresponding to each length, which pixels are proportionally distributed with respect to the line segment. Line of sight 106 when line of sight 106 intersects straight line 104
Since the spacing is equal to the size of one pixel, the ratio of the number of pixels represented by each line segment XI, X2, etc. to the corresponding spacing of one pixel in straight line 104 of screen AB is equal to lN5FAC. This lN5FAC corresponds to each line segment X1. of the straight line PQ. It may have different values depending on X2 etc.

Using the stored pixel intensity values of each straight line in the intermediate image, screen A is
The corresponding straight line 104 in B and all other lines for the final image can be obtained.

Other devices can be used when high resolution must be maintained as the object approaches the observer.

The device assumes that the input image is stored in a memory that can be addressed as a grid. The intersections of the grid are such points that the brightness of the input image is known.

Each output pixel can be projected into the input image grid.

The relative brightness of the output pixel is calculated from the area occupied by the projected output pixel. In FIG. 35, the projected output pixels are represented by dashed polygons.

The calculated brightness for the output pixel may be an area weighted average of the small region forming the projected output pixel.

This approach will be illustrated for a typical projected output pixel in FIG. In the figure, the rows of the input image grid are marked A.

B, C, D, columns are symbols a, b, c, d, e
It is expressed as. The small grid areas forming the inside of the projected output pixels are numbers 1, 2, 3.4, 5°6
．． It is expressed as 7,8.9. Each small region is surrounded by at least one of a solid line and a dashed line. For example, region 1 is bounded by the top and leftmost dashed lines, the bottom line of row A, and the right vertical line of column A.

The brightness of the output pixel is an area weighted average value. That is, the output - ε (brightness)) (area)) = 1 where the brightness ε (corner brightness for each small area) is the projected output pixel surrounded by the dashed polygon. is the area of a small region expressed as a percentage of the total area of .

The corner brightness for a small area is a convergent interpolation of the four corners of the square already formed in row A and column. The brightness at the lower right corner of region 1 is a convergent interpolation of the four corners of the square formed by row B and column C. The brightness at the lower left corner of region 1 is a convergent interpolation of the four corners of the square formed by rows B and columns. Those brightnesses for the corners of region 1 are averaged to give brightness j
(No. 1) is formed. This is multiplied by area j and becomes one of nine products that are added to calculate the brightness of the projected output pixel. In this way, the corner brightness for each small area is similarly interpolated and then averaged to yield a single value, which is then multiplied by the area of each area (the area is expressed as a percentage of the larger area). The products are then summed to calculate the brightness of the output pixel.The scene composition module considers the context of each object processed into a single scene or frame. It is something that should be properly placed.

A block diagram of this module is shown in FIG.

This diagram shows the sub-functions necessary to supplement all the overarching tasks required for these modular functions. The sub-functions are detailed below.

Channel combiner 80 forms the heart of the module. Data from multiple sources or channels is now combined on a pixel-to-pixel basis to form the final scene or image. Here, channel means video data. Of course, each source may be extended to include more than one color. As shown in FIG. 35, channel combiner 80 receives video from the surface and object channels. The channel combiner also receives range information for each displayed object from the object as well as the software controller of FIG. The channel combiner outputs the video data channels and trigger signals to a smoother 82 which performs the smoothing subfunction.

The channel binding supplement is shown in Figure 35.

The three basic elements are the object switch and the series-parallel boundary (s
/p-L/F) and the trigger generator 84 channel combination.

As shown in FIG. 35, the final object switch element 85 receives video data from two channel sources and range information from either the surface and object controller (via the series-parallel boundary) or the previous object switch. receive. The object switch then outputs the video channel selected on a pixel-to-pixel basis and the appropriate range of that channel. This selection criterion is referred to as "nearest occupied", where the video output is actually the video output of the closest object to the viewer that has a non-zero pixel bootara. One single range of values can be used to describe both two-dimensional and three-dimensional objects.

Object data is “field
zeroe+s). Each object switch in the array also outputs a "switch selection signal" which is an input to trigger generator 84.

As shown in FIG. 35, a trigger generator 84 receives switch selection signals from all object switches in the array. The output of the trigger generator 84 is the smoothing device 8 that controls the start of smoothing.
A single trigger signal used by 2. The characteristics of the signal generator are that (1) every change in the switch selection signal input that has an effect on the video output provided to the display device must cause a trigger signal for activation; The throughput delay caused by the number of pipeline stages in the generator must match the delay in the video path for the smoothing subfunction.

Scene value adjustment is performed by a scene value adjustment unit 86, which is used when scene width intensity correction is required. This correction includes day and night lighting, haze.

Applied to image compensation for rain, etc. These may be typical initial conditions or extremely slow changes in conditions. The regulator receives video data from channel combiner 80 and receives intensity correction values from the FOv computer. The output video from this subfunction is provided to the smoothing subfunction. The smoothing sub-function performed by smoothing unit 82 is utilized to simulate the edge characteristics of the sensor. Edges are attributed to adjacent pixels on the display line where the border lies. Such boundaries are trees caused by objects overlapping other objects, or boundaries that occur when transitioning from the background to an object. Although several edge smoothing algorithms are available, a 2 by 8 pixel process with Gax-Vixel weighting is preferred.

Special Effects The special effects functions performed by the special effects unit 12 are utilized to introduce translucent objects into the scene, such as smoke, haze, fog, or dust.

These effects also have a scope and can be introduced and rendered before or after other objects in the scene image.

FIG. 36 shows a block diagram regarding special effect functions.

Two sub-functions exist for this function, namely the series-parallel interface (vp
-VF) and special effects switch 88.

The S/P-I power performs its function by receiving serial data from the FOV computer and loading parallel range data into the range registers of the special effects switches.

The special effects switch 88 is very similar to the object switch, in that the nearest occupancy algorithm is used to select the range values to be sent to the next stage, and determines how the i-th and i+1-th video channels are handled. to be combined. The actual combination that affects the
It is a mathematical combination influenced by control and selection signals from the OV computer. The actual combination method is probably similar to the equation Y=ai+b(i+2) or bl+a(i+11).

[Brief explanation of drawings]

Figure 1 is a block diagram showing an overview of the CGS I system according to the present invention, Figure 2 is a system block diagram of the hardware used for database configuration, and Figure 3 is a method for dividing an object into sub-surfaces to obtain a three-dimensional image. Figures 4 and 5 are diagrams showing the process of applying high-fidelity objects and surfaces to optical disks that store object information in a database. Figures 6 and 7 are diagrams showing special Smoke treated as an effect object. A diagram showing the handling of dust, shadows, etc. and light transmission objects, Figure 8 is C
A diagram showing how to use a fan-shaped mask to generate luminous and flashing effects with the GS I system, Figure 9 is a CGS
IR images in the I database, Figure 10 is a diagram showing the flight path around a large object, Figures 11 to 14 are CGS I
Figure 15 is a diagram showing the absorption capacity of the system, Figure 15 is a schematic diagram of the visual field function, Figure 1-6 is a diagram illustrating the three-way image warping technique, and Figure 17 is an object from the database using the warpinder technique. Figure 18 is a block diagram showing the channels or pipelines for processing surface and special effects data; Figure 18 is a flowchart for implementing the strong control functions of the channel lock-up table of Figure 17; Figure 19 is a flowchart for implementing the warping technique; 20 is a block diagram of two same target processors that are used to process one input image in two directions. FIG.
FIG. 22 is a flowchart for the continuous interbleaching technique for the process shown in FIG. 6; FIG. Figures 24 to 26 are diagrams showing how image addresses are mapped to memory addresses in a spiral manner; Figure 27 is a diagram showing how images are spiral mapped to memory; Figures 28 and 29 are diagrams illustrating the process of the effect channel; Figures 28 and 29 are diagrams illustrating the horizontal and vertical mapping of object lines on the screen to the viewpoint; Figure 30 is a diagram illustrating the hardware pipeline that implements the equations in Table 3. , FIG. 31 shows the first pass vertical object line projection, FIG. 32 shows the second bus horizontal object line projection, FIG. 33 shows the output pixels corresponding to the alternating process algorithm, and FIG. 35 is a block diagram illustrating a screen construction module for compositing individual objects onto a single screen; FIG. FIG. 36 is a block diagram of special effect functions performed by the special effects unit 12 shown in the block diagram of FIG. 30... A/D card, 32... Access memory card, 34... Survey sheet card, 36...
・Y-axis twist card, 40... X-axis twist card, 56... memory control, 52... object investigation, 60... Y-axis processor, 62... X-axis processor, 64...Frame buffer, 70...
...frame buffer. Patent Applicant Honeywell Incorporated Sub-Agent Masaki Yamakawa (L/P) (1 person) FIG, 3 ・Shaku・FIG, q-' 1h FIG, 5 1F? 'I/4 E M/zrs) FIG, 6 8 η FIG, 7 FIG, gr-1 Yozan FIG, 12 A 8
CD E
FIG, I5 ×8εG XENDFIG,
IG 7, y74 ----- FIG, 23 512, + 512.2 512.3 512.4
... 512,512tLI umbrella r v, vt"y
7°゛Fl, 2'7 Difference 1') -1・壬PtLevel Desired FIG, 27 FIG, 28 FIG, 2'1 1t(y, 30 FIG, -'I FIG, 32 FIG, 36th Continued from page 1 Priority claim @July 30, 1982 Taoki United States (US)■
403386 @ July 30, 1982 @ United States (US) [Yes] 403
393 @ June 20, 1983 @ United States (US) ■ 505630 0 Inventor George W Ruessler 347 Drandview Avenue West, Roseville, Minnesota 55113 0 Inventor Michael O. Schrader United States 55418 St. Anthony, Minnesota 36T Avenue Northeast 3405 0 Inventor Dennis De Ferguson United States 55105 St. Paul, Minnesota Juliette Avenue 1947 Procedural Amendment Patent Application No. 14-04-2 Relationship with Case G: Lack of Patent Applicants: ]
Showa 5; Otsu year November 2-'j day fi'-= -
, -'d and the subject of amendment (1) Specification (-2) Drawings

Claims (1)

[Scope of Claims] (1) An observation plane within the coordinate system that has an observation point on the principal axis of a coordinate system, and an observation plane that extends perpendicularly to the principal axis and is separated from the observation point. A method of mapping an image, the method being applied to a rectangular digital input image arranged in columns and rows and represented as a matrix of pixel intensity values, said input image being arranged in columns and rows, said input image being mapped to said viewing plane. It is on the opposite side from the observation point,
forming an intermediate image on the input image; a first pass forming an intermediate image on the input image;
performing two passes perpendicular to a second pass forming an output image on the intermediate image; performing line transformation in the first pass;
a step of performing a line transformation in the pass of the line transformation, a step of performing calculation for each of the line transformations, a step of performing mapping, and a step of performing addition and averaging, converting said input image sequence into an intermediate image sequence having a length taken from said input image sequence such that there is a respective ratio defining the length of said intermediate image quality to said output image quality; converting intermediate image features into output image features on the viewing screen, having lengths obtained from each of the input image features, such that there is a respective ratio that determines the length of each input image feature; To make the output pixel of the nodame and the output pixel of the last imageable nodule,
The number of input pixels determined from the rows of the input image and the properties of the intermediate image is calculated based on the ratio, and the mapping process includes determining the number of input pixels for each line transformation. The process of mapping, summing and averaging the input pixels piece by piece to form a summation of the luminance values of the input pixels that can be added to each output pixel to obtain an average value thereof; A method for mapping an image to a viewing plane, characterized in that the composite values so obtained constitute each of said output pixels. (2) Rectangular pixel matrix of luminance value in a certain plane) I
In a method for linearly mapping digital input orthogonal images arranged in columns and rows represented as J-Knox onto an arbitrary convex quadrilateral, a first pass forming an intermediate image from said input image and a second pass for forming an output image from the intermediate image; an adjustment process; and a mapping process, wherein the first pass performs a transformation. further comprising transforming said input image columns into aligned intermediate vertical columns having lengths and positions determined from input image coordinates and said quadrilateral coordinates; A path further processes the intermediate image to perform a line transformation, thereby converting the properties into aligned horizontal output rows with lengths and positions determined from the corners of the intermediate image and the coordinates of the quadrilateral. the adjusting step: -adjusting the number of input pixels needed to create an output pixel; and the mapping step sequentially maps the input pixels to the output pixels, bit by bit, for each line transform. A method for linearly mapping a digital input orthogonal image onto an arbitrary convex quadrilateral, characterized by: (3) A method of mapping a digital input image to a viewing plane, the image comprising elements arranged in columns and rows representing a matrix of pixel intensity values in a 3D coordinate system, the coordinate system being an observation point on a principal axis and an observation plane spaced from the observation point and extending perpendicular to the axis, the coordinate system intersecting the principal axis and extending perpendicular to the observation plane; In a method of determining a reference plane and mapping a digital input image to an observation plane,
forming a first pass object line in the reference plane by sequentially projecting the rows of the input image to the reference plane in a perpendicular direction; simulating the projected divergent line passing through the observation plane and being separated from the origin and the corresponding first object line by an amount of at least one pixel, so as to draw an intermediate image line; processing each of said first pass object lines in a step of calculating the ratio of said first pass object lines to line segments; obtaining a pixel ratio for each first pass object line segment; mapping each row of pixels bit by bit to a corresponding first pass object line segment of the first pass object line; and said observation. said observing from said origin a diverging plane intersecting a plane to form a final image line and intersecting said input image to form a second pass object line and separated by at least one vixel; projecting a second path divergent line through a column of said viewing plane and said second path object line spaced from said origin by at least one pixel; processing each second pass object line by lengthening; and for each intermediate image property, processing each second pass object line segment and the last image line segment; obtaining a pixel ratio for a corresponding one; and mapping the pixels of a row to a corresponding second pass object line segment of the corresponding second path object line, step by step; has segments between the divergent lines having a length equal to the number of pixels separating the divergent lines, the second path divergent lines intersecting the viewing plane and extending horizontally; to form a final image line correctly positioned with a length equal to the number of pixels separating said second path divergent line;
A method for mapping a digital input image to a viewing surface, the method comprising: having a segment that intersects a path object line to form a segment. (4) A method of mapping within a 3D coordinate system having an observation plane extending perpendicular to the axis and separated from the observation point, the coordinate system including the principal axis,
having a reference plane extending perpendicular to the viewing plane, the method is applied to a digital input image represented as a matrix of pixel brightness values arranged in columns and rows;
A method for mapping in a 3D coordinate system, wherein the input image is on the opposite side of the observation plane from the observation point, and the method is applicable to map the input image to the observation plane, comprising: vertically projecting the columns one after another to the reference plane to form a first path object line in the reference plane; , sequentially processing each of the first pass object lines by irradiating radiation within the reference plane; assigning the total number of input image features; a step of sequentially mapping pixels to segments of the first pass object line of the corresponding first pass object line; an addition and averaging step; a step of processing the input image; and each of the second passes. a step of processing an object line; a second assigning step; a step of mapping row pixels; and a second summing and averaging step, wherein the ray intersects the observation plane. forming an intermediate image line having segments, each segment having a length of one pixel, each segment intersecting the first path object line to form segments of different lengths; is the total number of input image features in the first pass object line segment to obtain a vixel ratio of each first pass object line segment to a corresponding one of the intermediate image line segments.
The addition and averaging process adds the luminance values mapped to each of the first pass object lines to obtain an average value, and calculates the average value by adding the luminance values mapped to each of the first pass object lines, and calculates the average value by adding the luminance values mapped to each of the first pass object lines. In addition to line segments, the process of processing the input image includes extending a radial plane through the input image from an origin of the observation plane that is separated by one pixel. and each of the radiation planes intersects with the observation plane to form a final image and intersects with the input image to form a second path object line, and each of the second path object lines intersects with the input image to form a final image. Said process of processing said second pass ray by extending a second pass ray from said origin through said row of observation planes separated by one pixel and said second pass object line. processing each pass object line of 2, the second pass ray intersecting the observation plane to form a horizontally extending and properly positioned final image line; , assigning the total number of intermediate image properties to the second pass object line segment to obtain a pixel ratio for the second pass object line segment and the corresponding one of the line segments of the last image; The process of mapping vixels includes, for each of said intermediate image properties, mapping the pixels of a row to the second pass object line segment of the corresponding second pass object line bit by bit, and said second addition and averaging. For each segment of each second pass object line, the luminance values of each pixel mapped to it and which can be added to each final image pixel are added together, and the average is calculated. and give the composite value to the corresponding final image line segment,
A method for mapping in a 3D coordinate system, characterized by: (5) A method of mapping in a 3D coordinate system, the coordinate system comprising: an observation point on its principal axis; an observation plane extending perpendicular to said axis and spaced from said observation point; a reference plane including a principal axis and extending perpendicularly to said observation plane, represented as a matrix of pixel intensity values arranged in columns and rows, and arranged obliquely to at least one of said planes; A method of mapping in a 3D coordinate system, which may be applied to an elongated digital input image to map the input image to the viewing plane, comprising: making the columns of the input image orthogonal to the reference plane; a step of sequentially projecting a first path object line within the reference plane;
A ray forming an intermediate image line having segments each having a length of a vixel and intersecting the first path object line to form segments of different lengths is directed from the origin to the observation plane and the processing each said first pass object line by extending through a first pass object line; and a pixel ratio between the pass object line of said container 1 and a corresponding one of the segments of said intermediate image line. ' Assigning the total number of input image rows to the first pass object line segments in order to obtain '' for each input column the pixels of that input column in the corresponding first pass object line a step of sequentially mapping the segments of the first pass object line, and adding the luminance value of each pixel mapped to each segment of the first pass object line; averaging and applying composite values to intermediate, image line segments that intersect with said observation plane to form a final image line; and intersect with said input image to form a final image line; a line of the observation plane that separates a radial plane that forms a line and further intersects the input image to form a second path structure line from the origin by one pixel, and the input image; processing the input image by extending it through; projecting each structure line into a plane perpendicular to the observation plane and including the principal axis to form a second pass object line; (segment that intersects the observation plane and has a length of one pixel)? a second pass ray extending horizontally and correctly positioned to form a final image line, having a second pass ray extending horizontally and intersecting said second pass object line to form segments each having a different length from said origin. From, one
the second pass object line by extending through the second pass object line and the row of observation surfaces separated by a pixel;
processing each pass object line of the second pass object line to 1 to obtain a pixel ratio between each pass object line segment and a corresponding one of the line segments of the final image; assigning, to each intermediate image row, pixels of that row proportionally to segments of the second pass object line of the corresponding second pass object line; a step-by-step process of mapping the segments, and for each segment of each second pass object line, the luminance value of each pixel mapped to it, which can be added to each final image pixel; In addition,
A method for mapping within a 3D coordinate system, characterized in that the average is determined and the composite value is applied to the corresponding final image line segment. (6) A method of linearly mapping a digital input image, represented as a rectangular grid of luminance values in a plane, onto an arbitrary convex quadrilateral, the method comprising: projecting each output pixel within said quadrilateral onto said grid; and calculating the brightness of each output pixel from the area on the grid occupied by the output pixel; expressed as a rectangular grid of luminance values in a certain plane, characterized in that it is an area-weighted average value of the luminance values of all regions of the entire region and partial regions of the grid. A method for linearly mapping a digital input image into an arbitrary quadrilateral. (7) The pixels of said image are separated so that memory elements corresponding to any 2N consecutive pixels of any row and column of said image exist in different pixels and can be recalled simultaneously. (8) A method for mapping a two-axis square array image (D) to an address in memory, the method comprising mapping a two-axis square array image (D) to 2N memory banks that can be called by such that memory elements corresponding to any 2N consecutive pixels of a row and any column reside in different memory banks 1. and can be accessed simultaneously;
1. A method of mapping image addresses to addresses in memory, comprising the step of demultiplexing mapping of the vixels of a two-axis (AXA) square array image into 2N memory banks. (9) 2N independently recallable memory banks, row and column counter devices for horizontal or vertical address scanning of a two-axis (AXA) JIE rectangular array; a demultiplexing device for mapping pixels of said play into said banks such that any 2N consecutive pixels with any column are present in different banks and can be called simultaneously; 1. A circuit device for mapping image addresses of a two-axis (AXA) square array to memory addresses, comprising: (a defined area having a digital triaxial coordinate system of 1;
A library containing a plurality of two-dimensional image frames representing electromagnetic spectrum bands of two-dimensional images;
a simulation device for supplying observation data that determines the location and field of view of the scene recognition device with respect to the coordinate system of the defined area; an interface device for receiving the observation data; and the database. coupled to an interface device for determining which of the image frames are included in the field of view and their respective distances relative to the observed data and the frame data; a view processor device including a device for calculating the coordinates of a corner corresponding to the coordinates of a defined region of a corner of each image frame in the image frame; a control device including a device connected to the view processor and receiving data therefrom; a processing channel device controlled by the control device; and a scene composition device, wherein the defined region portion of the database includes scene composition data defining the location and dimensions of each frame relative to the coordinate system. , the control device is also connected to a library section of the database, and the data received from the viewing device is identical to the image frame contained in the viewing device and the data of the image frame included in the viewing device is the processing channel device is connected to the library portion of the database, the library portion of the database is connected to the control device and the coordinates of each of the included image frames; the channel device is operative to map the included frame into an enclosed space defined by a corresponding one of the coordinates of the corner;
The scene constructor determines the distance of the included frames such that the included frame closest to the scene recognizer includes one of the included frames that is more distant. computer-controlled image generation device responsive to the position and orientation of a scene recognition device, characterized in that it is connected to said processing channel for assembling an output scene on a digital basis. (11) A computer-controlled image that provides a raster display data for a rusk-type display that, in response to position data and orientation data of the scene recognition device, forms a two-dimensional perspective view of a portion of the three-dimensional device within the field of view of the scene recognition device. a database having a defined region section having a Nasital triaxial coordinate system and a library section having a plurality of two-dimensional rectangular image frames of image electromagnetic spectrum bands; and a coordinate system of the defined region. an interface device for receiving the observation data; an interface device connected to the interface device for retrieving the observation data; a controller connected to the library section of the database, the controller comprising: a visual field processor device connected to the local region section; a controller connected to the visual field processor device and including a device for receiving data calculated therefrom; at least one processing channel device to be controlled; and a scene composition device, wherein the defined region portion of the database defines the location and dimensions of each image frame relative to the coordinate system and the scale thereof. frame data; the processor unit determines which of the image frames are included in the field of view of the scene recognition device and their respective distances relative to the observation data and the frame data; and a device for calculating the coordinates of a corner of the screen corresponding to defined regional coordinates of a corner of each image frame included, the device receiving the calculated data of the control device includes the same image frames included in the field of view, the distance of the included frames, and the coordinates of each included image frame; The unit is connected to the control device and at least one of the processing channels, each of the at least one processing channel transmitting at least one of the included frames to a corresponding one of the corners of the screen. the scene constructor digitally assembles a scene based on the distances of the included frames, and the scene constructor is operable to digitally assemble a scene based on the distances of the included frames, A computer/event control image generation device characterized in that the included frame hides a frame that is further away than the included frame.