JPH02129769A - Scan converter - Google Patents

Abstract

PURPOSE:To efficiently attain the scan conversion of a polygon and the filing of an internal area by a constitution to be easily mounted by using two processing units, two common modules and a crossbar switch to constitute the scan converter. CONSTITUTION:The scan converter for finding out the inner point coordinates of a polygon in a raster image and processing them is constituted of two processing units 1, 2, a pair of data pass and memories 4, 5 and the crossbar switch 3 for alternatively switching connection between the units 1, 2 and the memories 4, 5. Consequently, processing time can be shortened, processing through-put can be improved and a substrate can be easily mounted.

Description

DETAILED DESCRIPTION OF THE INVENTION (7) Technical Field The present invention is in the field of computer graphics and relates to an apparatus for scanning and converting a polygon defined by vertex coordinates into a raster image.

Procedural Elements For C is a method for scanning and converting polygons represented in images.
computer Graphics"McGraw-H
There is an ordered edge list algorithm, as shown in ill (1gB5).

A raster image is an image in which pixels constituting an arbitrary image are arranged in a raster pattern on a screen divided into vertical and horizontal pixel columns.

A raster type is a scanning order in which a pixel set is scanned one line from left to right from the top row, and then descends to the next line and scans one line from left to right. Scanning that follows ζ in this order is called raster scanning.

A raster image is not just an arrangement of coordinates, but also the attributes of its pixels. The attributes include the color attributes of red R, green G, and blue B.

(B) Ordered Edge List Algorithm The ordered edge list algorithm is one method of scan conversion that allows the interior points of a polygon represented by a vertex coordinate sequence to be represented by a pixel set of a raster image.

The interior points of a polygon are the points inside the sides, and the points outside the sides are called the exterior.

To define a polygon, simply give the coordinates of the vertices. This is because a polygon is specified by sequentially connecting the vertices.

It is not immediately clear whether any point on the screen is inside or outside the polygon.

To define a polygon, it is sufficient to provide vertices, but in the field of image processing, it is not enough to simply define a polygon by drawing a straight line. Sometimes you want to give all the interior points a different color from the exterior.

In this way, filling the inside of a polygon with a certain color is called "
"Phil". This is the simplest image processing.

Furthermore, there are cases where the inside and outside of a polygon are clearly distinguished, and it is desired to change the internal color tone continuously in a certain direction.

For example, suppose that a certain color corresponds to a certain vertex, and a different color corresponds to an adjacent vertex. There are cases where it is desired to continuously change the hue on an edge connecting vertices.

This is called Gouro seeding.

Because of this requirement, the interior points of the polygon must be determined in advance.

An ordered edge list is a table that uses the X coordinate as an index and lists the first and last pixels of the pixel row determined by this X coordinate, which is the inner point of the image. . Since all pixels between these pixels are interior points, all interior points can be specified.

This will be explained using a simple example.

Consider an image as shown in FIG. This image is the vertex P1
It is a triangle with (4,1), P2 (2,8), and Pa (8-5).

The X coordinate should be pointing to the right, and the X coordinate should be pointing downward.

Pixels are counted from O.

The ordered edge list for ζ is shown in FIG.

The first vertical column on the left represents the X coordinate. The second column is the X coordinate of the first pixel that is the interior point at that X coordinate. The third column is the X coordinate of the last pixel that is the interior point at that X coordinate.

However, even if the X coordinate is fixed, the polygon may still be complex and have discrete sets of interior points.

In this case, the X coordinates of the starting point and ending point of a set of consecutive interior points are arranged.

In this case, the length of one consecutive set of interior points, that is, the length between the start point and end point is called the span.

In FIG. 4, for example, when 7=2, the pixel x=4 is the starting point of the interior point, and the pixel x=5 is the end point of the interior point set. Therefore, the edge list becomes 4 and 5.

For example, when y=3, the starting point of the interior point set is the pixel x = 4, and the ending point is the pixel x = 5. Therefore, the edge list becomes 4 and 6. In other words, among the 7=3 pixel sets, (4, 3), (513), and (stone, 3) are interior point sets.

A list with the X coordinate as a pointer as shown in Figure 5
It's called a packet.

Next, I will explain the process of creating an ordered edge list.

■ Initialize the y packet.

■ Select a pair of points that form the edges of the polygon from the vertex row. Let this be 5l(Xt,)'t), 52(x2.y2).

(2) Linear interpolation is performed between xl and Bresenham's algorithm, which will be explained later, can be applied to this.

■ Insert the found X intersection point into the appropriate position of the list so that it becomes a sorting class.

■ Perform steps from ■ to all edges of the polygon.

(1) Bresenham's 7) Lr Gorism S
1. It may seem easy because it calculates the X coordinate of the intersection of the straight line connecting S2 and the X coordinate, but it is not. A pixel is the smallest unit of an image. X must be an integer. An odd value of X is not allowed.

Bresenham's algorithm is a method for recursively obtaining linear interpolation values approximated by integers. The following formula % Formula % Xi is a value to be interpolated, and 6i is an error term that determines the interpolated value.

epig s 6mn5s xpis % Xmn- is a constant. It can be determined from the coordinate values of S□ and S2 according to the following equation.

e, B8=2(mod(dx,dy)-1) (
5e, , 8 = 2 mod (dx, dy)
(6xpJs=[dX/dy]+17 "mns""(dX/"I'J
(8dx == X2- Xl
(9dy-y2-yloo). Sl and S2 are the starting point and ending point of the line segment. 1y2 XI For the XI X coordinate values, find the integer value of the X intersection point. i is a number from Sl to S2. i
= x, 2 should not be confused with Xl, X2.

The initial value is eo = 2 (mod (dx, dy) -1)
Ql)xo==Xl
(It is 12.

[...] means to take the integer part of the value in this. When a decimal part is obtained during division, it is discarded.

iix, dy ha (9), (10) as shown, 2
This is the difference between the X coordinates of the two vertices. It should be noted that this is not a differential.

mod (dx, dy) is a symbol that indicates that of the value obtained by dividing dx by dy, the integer part is discarded and only the decimal part is taken. In other words, .

Since the algorithm of Br6s6nhzm is in the form of a recurrence formula, it is easy to handle and is suitable for calculation by computer.

This algorithm is, for example, “Procedural Elements For
Computer Graphics “McGraw”
-Hill 1985 “Fundamentals of' Interac
tive Computer Graphics''
J.D. FOLEY, A., VANDAM.

Learn more about Addison-Wesley 1982.

Although this is well known, it is difficult to understand that an edge ordered list can be created unless you understand its meaning, so I will briefly explain it.

FIG. 6 shows the screen. When the straight line 5IS2 and the X coordinate are determined, find the X coordinate which is an integer value. This is because X can only take integer values. The value of the center point of the pixel is (x
, y).

Let Ti be the intersection of this straight line and y=y0. Since the value of X in T1 is not an integer, the closest integer value Xi is taken.

The simplest way is to round off X, but then,
In some cases, the gradient becomes too steep or too gradual.

Therefore, we place emphasis on the fact that the slope does not change much. Let the value selected as the interpolation value be Xi, and the point (Xt*yt
) is written as Q.

T can be found immediately, but Qi cannot be found immediately.

The error 6i is defined as the difference between the X coordinates of Ti and Qi.

The slope of the straight line 5IS2 is dx/dy. However, this is generally not an integer. "The gradient of the straight line between i-1 and Ti is dx/dy. However, the gradient of the straight line connecting Q i -1 and Q must be an integer. If the difference in the X coordinates is 1, This is because the difference is an integer.

Therefore, the slope of the line segment connecting Qi-1 and Qt is dx/dy
is one of the integers between. Since the y increment is 1, the slope of the line segment is equal to the x increment. In other words, the increment of X is. As for which one to choose, it is decided to choose the one that reduces the absolute value of the error el.

For example, suppose that Q i-1 is larger than T1-1, that is, the error e is negative (8=T-Q').

To make it like I, Q increases from i-1 in the direction of W, and α
9 increases from Qi-1 in the direction of U.

In this case, the error increases as it increases in the U direction. As it increases in the direction of W, the error decreases. Therefore, let αX be the increment of X.

In other words, if %'i-1 is negative, add α to find Xi, and if ei-1 is positive (add 1 to find Xi).

As, from Xl-1, Xi is or is determined by. This is explained earlier (3), (4),
Equivalent to (7) and (8).

Now, the error e is T-Q, and a recurrence formula can be created for this as well.

The increase in X of the line segment 'r knee t'r is (dx/dy). However, compared to Xl-1, xiQ side has ei-
When 1 is positive, it increases by plm, so the error is (dx/
dy) increases by −Xpea.

That is, when ε1-1≧0, ε,=ε,−1+(”/d7)−xph(2t).

= nod (dx, dy) -1 (21J, which becomes equation (5) by multiplying by 2. Therefore, the definition of error is 2
By doubling it, equation (5) can be obtained. Since the error only needs to be determined to be positive or negative, it can be doubled or any number of times.

When the error is negative, X increases by xmns, so the increment of the error is the value obtained by subtracting this from (dx/dy).

This becomes nod(dx-dy).

B) Prior Art Using such an algorithm, an ordered edge list that defines the interior points of a polygon is obtained.

However, the scan conversion operation does not end there. Once the interior points are determined, a color is assigned to the interior points to distinguish them from the exterior. The draw may also be a ``fill'' operation that fills the interior points with a single color. In addition, Gouraud shading, which continuously changes the hue of interior points, is sometimes performed.

In any case, the color information of the interior points must be calculated and provided for each pixel.

This type of process requires two types of operations when it comes to scan conversion.

Furthermore, there are many polygons within one screen. An ordered edge list must be found for each polygon and color values must be assigned to the interior points. To find the interior point coordinates of a polygon in a raster image, it can be processed by a single processor or by two processors, one that creates an ordered edge list and the other that calculates the interior point coordinates from the ordered edge list. Processing was pipelined between the processors that generated it.Even in this case, the connection between the processors was simply shared memory.

(3) Problems to be Solved by the Invention When processed by a single processor, the rask transformation takes too long.

In a system configured with two processors and one shared memory, while one processing unit is accessing the shared memory, the other processing unit is in a wait state. This reduces processing throughput.

Some devices are configured to selectively connect two processing units and two shared memories using a crossbar switch. This is x, y, z.

As the number of scan axes such as r, g, and b increases, the data path width to be switched increases. This makes board mounting difficult. There was a problem like this.

ψ) Configuration The scan conversion device of the present invention includes two processing units, a pair of data paths and memory shared by these, and a crossbar switch that selectively switches the connection between the processing units and the data path/memory. It becomes more.

The schematic configuration of the scan conversion device of the present invention will be explained with reference to FIG.

The processing unit 1.2 is a processing unit that independently performs specific processing.

The data path and memory modules 4.5 are identical but can be shared by each processing unit 1.2.

The crossbar switch 3 is a processing unit 1.2
and the shared module 4.5 are alternatively connected.

When connected like A-G, B-D, and when connected like A-D
, B-C.

The term "crossbar switch" is used to refer to a mechanical switch that switches telephone exchanges6.However, here, it does not mean that; instead, it is used to refer to a pair of n buses and n buses in any combination. It is used in the sense of a means of connecting to one.

The shared data path and memory module 4.5 further includes an interpolator 6, an interpolator 7 and an interpolator 8.

As will be described later, interpolator 6 is an X interpolator, interpolator 7 .
8 can be made into a z1r1g1b interpolator Y.: Can be done.

Memory 9.10 is further included in shared module 4.5. Memory 9 is accessible by interpolator 6 and processing unit. Memory 10 is a memory local to the interpolator 7.8. Local means that data can be exchanged with the interpolator 7.8, but not from the processing unit.

The shared memory 4.5 further includes a three-state buffer 11.

(4) As an example of operation, consider scan conversion between a two-dimensional coordinate space and a one-dimensional color space.

Assume that the crossbar switch 3 in FIG. 1 is in the connected state of AC and BD.

The data of the polygon's vertex coordinates
．． 7.8 is sent for each vertex.

Interpolator 6.7.8 holds data for the previous vertex. The interpolator sequentially generates values obtained by performing linear interpolation between the value of the immediately preceding vertex and the value of the newly sent vertex using an integer value by the difference Δy in the y coordinate according to Bresenham's algorithm.

As explained earlier, in order to calculate Bresenham's recurrence formula, it is necessary to calculate a constant term, which requires a divider to obtain a quotient and a remainder.

Also, an adder is required to calculate the recurrence formula.

The interpolator includes the necessary dividers and adders.

What was previously described was an algorithm for finding the X coordinate of the interior points of a polygon. However, the data of the three primary colors on the sides of a polygon can also be found using this algorithm. The interpolator for the X coordinate and the interpolators for the three primary colors r1g and b are similar.

However, how the color is interpolated differs depending on whether it is fill or Gouraud shading. This is, however, a matter of what to do with the color of the vertices, and given the color of the vertices, it can be interpolated using the same algorithm.

The interpolated X value is sent to memory 9 and stored.

The interpolated yl_C value is sent to memory 10 and stored.

The memory 9 stores the y packet, the X intersection L and other intersection data (here y, C) in the ordered edge list.
, the address of memory 10 is stored.

Intersection data other than the X intersection is stored at the designated address in the memory 10.

The above operations are performed for all vertices that make up the polygon.

Then, the final ordered edge list is completed in memory 9.10.

Here, memory 9.10 means crossbar switch 3
This is the one in the shared module 4 connected to the 0 point. Connected to A-C, processing unit 1
This is because the shared module 4 was used to create an ordered edge list.

At this point, switch the crossbar switch 3.

The connection will be A-Dl, B-C.

The processing unit 1 then uses the shared module 5 on the right side to perform the same processing on the next polygon.

On the other hand, the processing unit 2 searches the ordered edge list stored in the memory 9.10 and sets the data of "both end points of the span" in the interpolator 6.7.8.

As already explained, a span is a set of consecutive interior points when cut at a certain X coordinate. Since the start and end points of the interior point set are arranged in an ordered edge list, when two elements from the beginning are made into pairs in the ordered edge list, the line segment represented by the pair is called a span. .

The interpolator 6.7.8 interpolates data for each difference in the X direction from the data at both end points of the span. This interpolation is also Bres
Enham's algorithm can be used.

The interpolated data passes through the 3-state buffer 11 and is output to the outside. At this time, the 3-state buffer 11 of the other shared module 5 is in a high impedance state. Therefore, no data is output from the shared module 5. In other words, one of the three-state buffers in module 4.5 is placed in a high impedance state so as not to interfere with the data output of the other.

It is a kind of multiplexer.

During this time, the processing unit 1 and module 5 are creating an ordered edge list for the next polygon ζ.

The above processes of the processing unit 1.2 are carried out in parallel and simultaneously.

When both processes are completed, the crossbar switch 3 is switched. Similar processing is then repeated for new polygon data.

At the beginning of this iteration, the ordered edge list must be initialized to an empty state.

To do this, initialize the y packet. For this reason, it is only necessary to provide a means for instantaneously clearing the memory 9 that holds the y packet. This will speed up the processing.

In this way, processing by the processing unit 1 that creates the ordered edge list and the processing unit 2 that creates points inside the polygon from the ordered edge list are performed simultaneously. This allows processing throughput to be increased.

Furthermore, by sharing the data path of the interpolators 6, 7, and 8, the number of terminals of the crossbar switch 3 can be reduced.

Therefore, board mounting can be facilitated.

(R) Embodiment FIG. 2 shows an embodiment of the present invention. They correspond to the processing unit 1.2 and are labeled A and B.

The processing unit includes a sequencer 12, a control memory 13, a pipeline register i4, an ALU 15, and the like.

The crossbar switch 3 has four ports A, B, C, and D. Connection A-01B-D, A-D, B
-C connection can be switched.

Each port) A, B, and C1D consists of an input-only terminal, an output-only terminal, and an input/output terminal.

4.5 is a shared module that can be accessed by both processing units A1 and B.

FIG. 3 shows the contents of the data portion of the shared module.

This includes the X interpolator 21.2 interpolator 22, r interpolator 23,
g interpolator 24, b interpolator 25, y register 26, address register 27.28, buffer 29, memory 30,
It consists of a memory 31, a buffer 32, etc.

The X interpolator 21 interpolates the X coordinate value.

2 interpolator 22 interpolates two coordinate values.

The r interpolator 23 is for interpolating red color component values.

The g interpolator 24 interpolates the green color component values.

The b interpolator 25 interpolates the blue color component value.

As for the interpolation method, the already mentioned Bresenham algorithm can be used as is. Instead of the X coordinate,
The color component values at both ends of the line segment 5IS2 are used as data. For r interpolation, the r component in Sl is rl, and the r component in S2 is r
2, interpolate between them. Interpolation is performed first in the X direction, and then in the X direction.

The y register 26 is a register that holds the value of the X coordinate.

The address register 27 is a register that holds the access address of the memory 30.

The address register 28 is a register that holds the access address of the memory 31.

Buffer 29 constitutes a data flow for writing data in address register 28 to memory 30 .

Memory 30 corresponds to memory 9 in FIG. This is the y packet, the value of the X intersection, and other intersection data z1 rlgl
This is a memory that holds the address of the memory 31 where data b is stored. The address of the corresponding data in the memory 31 is called a pointer, and in order to put this into the memory 30, the address register 28 is branched and the buffer 2
9, the data is taken into the memory 30 as data.

The memory 31 holds interpolated values of z, r%g1b with addresses attached thereto.

The value of X% ``s l gs b'' as a result of interpolation and the value of y, which is a constant, are outputted to the outside through the buffer 32.

When performing polygon scan conversion, "filling" means filling in the internal area on the raster image.

This can be done by keeping the color component values r, g, and b constant for the interior points.

In the case of Gouraud shading, the color changes continuously from vertex to vertex. This can be determined from the color data of the vertices using the following procedure.

(1) First, interpolate the colors at both ends of the edge in the y-axis direction on the edge.

This is the same procedure as the procedure for interpolating the X coordinate values at both ends of edges in the y-axis direction when creating an ordered edge list. However, the difference is that not only the X coordinate but also color data is added to the elements of the ordered edge list.

(2) Interpolate the colors at both ends of the span in the ordered edge list in the X-axis direction. In this way, the color value of the point inside the polygon is obtained.

f) Effects This device can efficiently perform scan conversion of polygons and fill internal areas using two processors (and is easy to implement), especially when using an ordered edge list algorithm.

In fields such as computer graphics, it is effective when used for polygonal fill and Gouraud shading.

[Brief explanation of drawings]

FIG. 1 is a configuration diagram of a scan conversion device of the present invention. FIG. 2 is a configuration diagram showing a processing unit according to the embodiment. FIG. 3 is a configuration diagram of the shared memory. 1.2...Processing unit 3...
......Crossbar switch 4.5...Data path and memory modules 6.7, 8...Interpolator 9...Memory 10... ...Memory 11...3-state buffer 12...
...... Sequencer 13 ...... Control memory 14 ...... Vibration line register 15 ...
......ALU 21......X interpolator 22...2 interpolator 23...r interpolator 24... ......g interpolator 25...b interpolator 26...y register 27.28...address register 29...
...Buffer 30 ...Memory 31 ...Memory 32 ...Buffer inventor Yamashita

Claims (3)

[Claims] (1) A scan conversion device for displaying the internal area of a polygon defined by the coordinates of vertices as a different color from the outside on a screen divided into a large number of pixels lined up vertically and horizontally, which uses two processing methods. Units 1 and 2, and two data path and memory sharing modules 4 and 5 that are alternatively shared by these processing units 1 and 2.
and a crossbar switch 3 that selectively switches the connection between the two processing units 1 and 2 and the two shared modules 4 and 5, and the data path of the shared modules 4 and 5 includes a divider and an adder for calculating the quotient. One processing unit and one shared module calculate the x-coordinate of the line segment that separates the polygon from the coordinates of the polygon's vertices for each y-coordinate. Create an ordered edge list and define the other processing unit and the other shared module.
A scan conversion device characterized in that the interior points of a polygon and the color component values for the interior points are determined from the determined ordered edge list. (2) The scan conversion device according to claim (1), wherein the memories of the shared modules 4 and 5 are comprised of a portion local to the shared data path and a portion accessible from the processing unit. . (3) The memory portion included in the shared modules 4 and 5 and accessible from the processing unit is provided with means for instantaneously clearing the contents of the memory. 2) The scan conversion device described in section 2).