KR20050013287A - method for filtering of triangle strip in rendering system of view-dependency Level Of Detail based - Google Patents

Abstract

PURPOSE: A method for filtering a triangle strip in a view-dependent LOD(Level Of Detail)-based rendering system is provided to effectively remove a repetitive apex except a swap calculation, thereby effectively rendering a triangle mesh. CONSTITUTION: The triangular mesh constructed by a triangle in a three dimensional graphic image is expressed by a triangle strip for performing an efficient rendering(S10). A simplified operator such as an edge collapse in the triangle mesh is used to two apexes of the triangle, thereby simplifying the triangle mesh(S20). The triangle strip of the original mesh are replaced with an activated ancestor. The activation is an apex shown by the given view-dependent condition(S30). Apexes for forming a triangle are extracted from the triangle strip updated by a skip strip through a DTFF(Degenerate Triangle Free-Filter) and the apexes are optimized by using the extracted efficient triangle(S40). An efficient triangle is detected by the DTFF and the apexes optimized by the triangle strip is transmitted to a graphic engine, thereby performing the rendering(S50).

Description

Method for filtering of triangle strip in rendering system of view-dependency Level Of Detail based}

TECHNICAL FIELD The present invention relates to a three-dimensional graphics accelerator, and more particularly, to an improved triangular strip filtering method for skip strip structures for rendering triangular meshes effectively in a view-dependent LOD based rendering system.

Recently, 3D graphics techniques have been introduced to the PC, and 3D graphics hardware, which can efficiently handle a large number of polygons and special effects such as light effects, has become the mainstream of the market in order to construct a more realistic and colorful screen. .

In particular, as games that provide high realism to users using 3D screens have received great response, research on this has made great progress.

Such 3D graphic images are mainly composed of points, lines, and polygons, and most 3D rendering processes mainly have a structure of processing triangles at high speed.

As described above, the 3D graphic image may be represented by a triangle mesh, which is a set of triangles composed of triangles. For efficient rendering, the 3D graphic image is formed by forming a series of triangular strips instead of independently processing triangles constituting the triangle mesh. Triangular strip structures have been proposed. In addition, a skip strip structure has been proposed as a representative data structure for using the triangle strip structure for efficient rendering of the triangle mesh in a view-dependent LOD based rendering system.

In general, for view-dependent substrate rendering, a segmentation operator, such as edge collapse, is applied to two vertices, and these two vertices are represented by a vertex hierarchy tree, usually represented by a parent-child relationship.

And the triangular strip shares the sides of two consecutive triangles except the triangle starting as one of the data structures providing a concise representation of the triangular mesh.

In other words, each triangle in the mesh is rendered by sending three vertices separately to the graphics engine. If you render triangles independently to render T triangles in a normal regular mesh, 3T vertices are sent to the graphics engine. .

However, the triangular strip reuses two vertices of the previous triangle in all but the first triangle of each strip, so that only T + 2 vertices are sent to the graphics engine. Thus, the number of vertex iterations and rendering time is reduced by approximately 1/2 to 1/3.

These triangular strips are currently supported directly by graphics hardware and standard graphics libraries such as OpenGL.

1 (a) and (b) show two types of triangular strips according to the prior art, a sequential strip and a generalized strip.

In the sequential strip shown in Fig. 1A, the strip forming directions are alternately repeated in the clockwise direction CW and the counterclockwise direction CCW. On the other hand, the generalized strip shown in FIG. 1 (b) may not repeat the direction of strip formation. Therefore, a swap operation is required to convert the generalized strip into a sequential strip.

This swap operation reverses the order of the two previous vertices, implemented by iterating over them.

That is, by repeating the vertex 'c' in FIG. 1 (b), 'cdc' between the triangle 'bcd' and the triangle 'dcf' is generated.

However, the repetition of these vertices produces degenerate triangles.

In the simplification process, where two vertices are pre-condensed, the triangular strip replaces the shortened and dissipated vertices with the abbreviated vertex (parent).

Like the view-dependent LOD-based rendering system, in order to use the rendering efficiency of the triangle strip structure when the phase information of the triangle mesh changes every frame, the skip strip structure representing the triangle strip structure that is changed by the prediction operator is used. need.

In the skip strip technique, the highest resolution triangular mesh is represented as a triangular strip, and a modified triangular strip is substituted to the graphics engine by substituting parent nodes for child nodes appearing in the triangular strip at a given stage to render a given LOD mesh.

In this case, the nodes extracted from the skip strip are repeatedly accumulated and accumulated at a lower resolution, and these vertices form the degenerate triangle having a zero size.

The increase in the degenerate triangle not only increases the size of the graphic data but also increases the complexity of the data, thereby adversely affecting the performance of the graphic engine.

Therefore, such a degenerate triangle is a triangle that does not contribute to the final screen and a technique for effectively filtering the sequence of these vertices is required.

At present, a simple triangle strip scanner (STSS) has been proposed to solve this problem. El-Sana, F. Evans, S. Skiena, and E. Azanli, Efficiently Computing and Updating Triangle Strips for Real-Time Rendering '', The Journal Computer-Aided Design , vol 32, no 13, 2000, pp. 753-772.]

The STSS detects a vertex pattern of (aa) ⁺ and (ab) ⁺ from the skip strip node in such a way that it finds a sequence of repeated vertices in the display strip and does not render these vertices by rendering them (aa) and (ab) Replace each with. The modified display strip is then sent to the graphics engine.

However, the STSS removes the degenerate triangle well when the model level is high, but is still inefficient when the level is low. Because there is no way to remove degenerate triangles of patterns such as (aabb) and (abbc), these triangles are more likely to occur as the model goes to lower levels.

Accordingly, an object of the present invention is to provide an efficient triangular strip filtering method constituting a triangular strip by efficiently removing the repetitive vertices except for a swap operation.

Another object of the present invention is to provide a triangular strip filtering method that can reduce the number of advantages sent to the graphics engine by using a degenerate triangle free-filter (DTFF).

1 (a) (b) show two types of triangular strips according to the prior art, a sequential strip and a generalized strip;

2 is a flowchart illustrating a method of filtering triangle strips in a view-dependent LOD based rendering system according to the present invention.

3 (a) is an embodiment of a sequence for scanning vertices from a regular strip node in accordance with the present invention.

3 (b) is an embodiment of a sequence for scanning vertices from a swap strip node according to the present invention.

4 is an embodiment of a sequence for scanning vertices from a skip strip node in accordance with the present invention.

5 (a) to 5 (h) show various models using the method of filtering a triangular strip according to the present invention in an experiment.

6A, 6B, 6C, and 6D are graphs showing the number of vertices sent to the graphics engine at various LOD levels.

In the three-dimensional graphics accelerator according to the present invention for achieving the above object, the feature of the filtering method of the triangular strip is the step of expressing a triangular mesh consisting of triangles in a three-dimensional graphics image as a triangular strip, the generated triangular mesh Applying a simplification process such as an edge collapse operator to the two vertices of the triangle, expressing the simplification process as a skip strip, and degenerating a triangle strip from the skip strip Extracting only vertices that form a triangle through the Degenerate Triangle Free-Filter (DTFF), optimizing the vertices with a triangular strip using the extracted valid triangles, and graphicizing the vertices optimized with a triangular strip using the valid triangles. Send it to the engine to perform the rendering It consists of including.

In this case, reducing the repeating vertex through the DTFF may include scanning a sequence of vertices from the skip strip, extracting only vertices and directions forming a triangle, and successive patterns of two neighboring triangles through the extracted vertices and directions. And a step of irradiating with each other, and if the directions of successive patterns of the triangle vertices are different from each other and share sides of two consecutive triangles except the starting triangle, performing regular processing; If the continuous pattern of triangle vertices is the same in direction and shares two consecutive triangles except for the starting triangle, performing a swap process; If the typical pattern is other than the above two cases, it is preferable to include the step of performing a reset process.

In addition, the reset process may include detecting two vertices of a triangle in which a common side of neighboring triangles does not exist when forming a strip of the triangle strip, implementing a triangle strip including up to a triangle having the detected two vertices; And detecting two vertices of a triangle having no common side of the triangle from a triangle including a vertex shared with one of the two vertices detected, and repeating the above steps. .

In addition, the reset process preferably ends with one strip by one triangle and another strip by another triangle.

Other objects, features and advantages of the present invention will become apparent from the following detailed description of embodiments with reference to the accompanying drawings.

A preferred embodiment of the triangular strip filtering method in the 3D graphic accelerator according to the present invention will be described with reference to the accompanying drawings.

2 is a flowchart illustrating a method of filtering triangular strips in a three-dimensional graphics accelerator according to the present invention.

Referring to FIG. 2, first, a triangular mesh composed of triangles in a 3D graphic image is represented by a triangular strip for efficient rendering (S10).

In addition, a triangulation operator such as edge collapse is applied to the two vertices of the triangle in the triangular mesh to simplify the triangular mesh (S20).

In this case, two simplified vertices have a relationship between a shortened child vertex and a shortened parent vertex, and the parent-child relationship according to the simplified representation is represented by a skip strip structure.

Thus, the triangular strip of the original mesh is replaced with each of the simplified vertices closest to the active ancestor to render the mesh to the strip at a given LOD level. In this case, the activation is a vertex seen by a given view-dependent condition (S30).

The triangular strip in the skip strip has two forms for rendering.

One is a strip for the original model, and the other is a strip for displays that fit the LOD level.

The display strip is extracted from the skip strip node in the rendering stage, which has the geometric information of the vertices and the information of the parent and the child list. At this time, exactly one child pointer is activated.

Also, in the simplification process, if a node is collapsed as a parent node, the node's parent pointer is marked as active, otherwise it is marked as inactive.

If the node's parent is active, continue following until the inactive parent is found. In the current step, the node is shown by the deactivated pointer.

However, as the level of the model decreases, repeated vertices appear several times in the display strip. This will increase the number of vertices sent to the graphics engine.

Therefore, the triangle strip updated through the skip strip is extracted only through vertices forming a triangle through a degenerate triangle free-filter (DTFF), and the vertices are optimized by using the extracted valid triangle. (S40).

In more detail, the DTFF first scans the sequence of vertices from the skip strip to extract only the vertices and directions that form a triangle.

Then, through the extracted vertices and directions, a continuous pattern of two neighboring triangles is examined.

As a result of the above investigation, when the direction of the continuous pattern of the triangle vertices is different from each other as shown in FIG. 3 (a), and the triangles are 'abc' and 'def', Perform regular processing to add a sequence of 'abcf' to the display strip. As a result, the triangle 'def' reduces two vertices, because the common side (b, c) is reused.

In this case, the regular process detects two vertices of a common side of a neighboring triangle in which the strip forming direction of the triangular strip is alternately repeated in the clockwise direction and the counterclockwise direction, and implements the triangular strip reusing the detected two vertices. Is done.

As a result of the investigation, as shown in FIG. 3 (b), when the continuous patterns of the triangle vertices have the same direction and the triangles are 'abc' and 'def', "(b = e) AND (c = d)" If so, a swap process is performed to add the sequence of 'abcbf' to the display strip. At this time, since vertex b is repeated due to the swap operation, the triangle 'def' reduces one vertex.

In this case, the swap process detects two vertices of neighboring triangles in which the strip forming direction of the triangular strip is not repeated in the clockwise direction and the counterclockwise direction, and repeats the first vertex of the detected two vertices once. The process consists of implementing the strip.

In addition, as a result of the investigation, if the continuous pattern of the triangle vertices as shown in FIG. 4 is not the above two cases, a reset process is performed to form a new display strip. In other words, the current display strip ends with a triangle 'abc' and again creates a new display strip 'def'.

As described above, valid triangles are detected through the degenerate triangle free-filter (DTFF), and vertices optimized to triangle strips are sent to the graphics engine for rendering.

The triangular strip filtering method according to the present invention as described above will be described in detail by way of examples.

4 is an embodiment of a sequence for scanning vertices from a skip strip node in accordance with the present invention.

When the skip strip is performed with reference to FIG. 4, the conventional STSS obtains a vertex sequence of 615 515 502 615 206 620 620 580 66 761 761 258 761 as a display strip.

DTFF, on the other hand, gets 502 615 206 620 (-1) 620 580 66 761, where (-1) points to the beginning of a new triangular strip and is not rendered. Therefore, after rendering up to 502 615 206 620, rendering up to 620 580 66 761 is performed again.

In this embodiment, it can be seen that the number of vertices transmitted to the graphics engine is 13 for STSS and 8 for DTFF, and 12 for transmitting vertices individually.

Model T V VTS TS S Bishop 496 250 570 One 72 Eight 1536 766 1648 24 64 Femur 7798 3897 10254 237 1982 Triceratops 5660 2832 7372 144 1424

Table 1 shows the three-dimensional data modules used in the experiment. In this table, T represents the number of triangles, V represents the number of vertices, and VTS represents the number of vertices in the strip. TS is the number of triangular strips, and S is the number of swap operations.

Each three-dimensional model was stripped by STRIPE 2.0.

In addition, a simplification process was performed with half-edge collapse to construct a skip strip, and the metric used a quadratic error metric (QEM).

The reason for using the semi-decay operation as a simplification is that you can reuse the geometry information for the vertices of the previous level. This has the advantage of reducing storage space when the level of the model changes.

QEMs are also well known for their speed, accuracy and ease of implementation.

5 (a) to 5 (h) are diagrams showing various models using the filtering method of the triangular strip according to the present invention in an experiment. 5 (a) to 5 (h) uniformly extract the sequence of vertices in the skip strip node, and do not limit the viewpoint dependence.

Such extraction does not affect the filtering algorithm and provides the generality of continuous LOD.

In the experiment, the number of vertices sent to the graphics engine was calculated for DTFF and STSS, respectively, using four three-dimensional models at various LOD levels. We also compared the number of vertices and the DTFF and STSS of the triangles individually transmitted (TETI).

6A, 6B, 6C, and 6D are graphs showing the number of vertices sent to the graphics engine at various LOD levels.

As shown in Figures 6A, 6B, 6C, and 6D, the experimental results show that DTFF can reduce the number of vertices significantly greater than STSS, especially as the model goes to a lower level.

At this time, the DTFF also performed better than TETI at the overall level of the model.

On the other hand, the STSS reduced the peak of DTFF when the model level was high, but the number of vertices was higher than that of TETI at the low level.

Table 2 shows the rate of vertex reduction sent for rendering.

LODs Bishop Eight Femur Triceratops DvS DvT SvT DvS DvT SvT DvS DvT SvT DvS DvT SvT 10 73.2 36.8 -136.1 52.7 18.4 -72.4 57.0 10.1 -109.3 56.8 7.2 -114.6 15 72.2 43.2 -104.5 44.4 27.8 -29.9 47.8 15.2 -62.3 46.9 11.6 -66.6 20 68.2 46.9 -66.7 38.8 32.0 -11.1 41.6 21.1 -35.0 40.6 17.5 -39.0 25 58.1 42.7 -36.6 35.0 34.1 -1.4 36.5 25.0 -18.1 35.2 22.2 -20.1 30 53.2 44.4 -18.9 30.0 39.9 14.1 31.8 28.5 -4.7 30.0 25.8 -6.0 35 45.5 44.2 -2.3 26.0 42.7 22.5 27.8 31.8 5.5 25.1 28.8 409 40 42.9 48.8 10.4 23.7 45.9 29.1 23.8 34.5 14.1 21.5 32.4 13.8 50 36.0 53.2 26.9 16.0 50.3 40.9 17.4 39.3 26.4 14.9 37.7 26.9 60 29.1 55.6 37.4 11.0 54.5 48.8 12.4 43.7 35.7 10.2 42.4 35.8 70 22.4 58.3 46.2 6.7 58.6 55.7 8.4 47.5 42.7 6.8 46.7 42.8 80 12.5 58.9 53.0 4.1 60.3 58.6 5.5 50.9 48.0 4.1 50.6 48.4 90 3.9 59.6 58.0 1.5 61.7 61.2 2.4 53.8 52.7 1.6 53.8 53.1 100 0.0 61.7 61.7 0.0 64.2 64.2 0.0 56.2 56.2 0.0 56.6 56.6 AVG 33.6 52.2 9.8 17.8 49.0 31.3 19.4 38.9 15.5 17.9 37.3 14.7

As shown in Table 1, it can be seen that the DTFF reduced the number of vertices of about 44.4% and 22.2% than TETI and STSS on average.

As such, the present invention proposes an efficient filtering algorithm based on view-dependent LOD rendering that is frequently used in continuous LOD, and the core idea of DTFF is to share common sides of consecutive triangles in a sequence of vertices obtained from skip strip nodes during rendering. As a result, the performance is better than TETI at the overall level, and the lower the level of the model, the better than the STSS.

In addition, the DTFF is easy to implement and fits well with existing skip strip approaches.

Those skilled in the art will appreciate that various changes and modifications can be made without departing from the spirit of the present invention.

Therefore, the technical scope of the present invention should not be limited to the contents described in the embodiments, but should be defined by the claims.

In the three-dimensional graphics accelerator according to the present invention as described above, the filtering method of the triangular strip can reduce the number of advantages that the DTFF has to the graphics engine as compared to the STSS as the model goes to a lower level. Sending each triangle individually shows great performance. In other words, compared to STSS, DTFF can reduce the number of vertices from 73.2% to 22.2% on average.

Claims (5)

Representing a triangle mesh consisting of triangles in a three-dimensional graphic image as a triangle strip, Applying a simplification process such as an edge collapse operator to the two vertices of the triangle for the generated triangle mesh; Expressing the simplified process in a skip strip; Extracting only the vertices that form a triangle through a degenerate triangle free-filter (DTFF) from the skipped strip, and optimizing the vertices with a triangle strip using the extracted valid triangles; And sending the vertices optimized to the triangular strip to the graphics engine using the valid triangle to perform rendering. The method of claim 1, wherein reducing the iteration vertex through the DTFF Scanning the sequence of vertices from the skip strip to extract only vertices and directions that form a triangle; Investigating a continuous pattern of two neighboring triangles through the extracted vertices and directions; Performing a regular process if the directions of the continuous patterns of the triangle vertices are different from each other and share sides of two consecutive triangles except the starting triangle; Performing a swap process if the consecutive patterns of the triangle vertices have the same direction and share the sides of two consecutive triangles except the starting triangle; And, if the continuous pattern of the triangle vertices is other than the above two cases, performing the reset process. The method of claim 2, wherein the regular treatment Detecting two vertices of a common side of a neighboring triangle in which the strip forming direction of the triangular strip is alternately repeated clockwise and counterclockwise; And implementing a triangular strip for reusing the detected two vertices. The method of claim 2, wherein the swap process is Detecting two vertices of a neighboring triangle in which the stripping direction of the triangular strip does not repeat alternately clockwise and counterclockwise; And implementing a triangular strip that repeats the first vertex one of the detected two vertices once. The method of claim 2, wherein the reset process Detecting triangles when there are no common sides of neighboring triangles in the formation of the strip of triangle strips, Terminating one triangle strip by the detected one triangle; And a new triangle strip is started by another triangle.