JPH03139774A - Aliasing removal system for image formation - Google Patents

Abstract

PURPOSE:To remove aliasing without making the whole blur nor increasing the cost by detecting an area which has discontinuity i the depth direction by applying a differential filter to a Z buffer, and filtering data only in the discontinuous area. CONSTITUTION:A image generation part 11 outputs color information on an image, generated on the Z buffer 12 stored with depth information, to a frame buffer 13, where the information is stored. The differential filter 14 differentiates an image of the depth information stored in the Z buffer 12 and checks whether the differentiated value as the differentiation result is larger than a previously set threshold value or not. Then it is judged that the depth is discontinuous when the differentiated value is larger than the threshold value and continuous when not, and an averaging filter 16 filters only a part where the depth is discontinuous on the frame buffer 13. Consequently, the aliasing is removed without making the whole blur nor increasing the cost.

Description

[Detailed Description of the Invention] [Summary] An object of the present invention is to provide an aliasing removal method for image generation that can remove aliasing without blurring the entire image and without increasing calculation cost. Then, referring to the contents of the Z buffer and frame buffer obtained as a result of image generation, a differential filter is applied to the Z buffer to find the area corresponding to the pixel with depth discontinuity among the pixels on the frame buffer. It is configured to detect, apply an averaging filter to the detected area, and replace the resulting average value of neighboring pixels as a new pixel value.

[Industrial Application Field] The present invention relates to an aliasing removal method in image generation.

With the increase in speed and capacity of computers in recent years, the application fields of image generation by computers are expanding to various design, art,
His work covers a wide range of areas, including commercial films and animations, the creation of various presentation materials, and the visualization of various simulations. Generally, in these application fields, there is a tendency to pursue realism, and it is important to see how realistic the generated image looks.

[Prior Art] In such conventional image generation, a problem is the occurrence of aliasing.

Aliasing, also called jaggies, is a phenomenon in which the outline of an object or the pattern on the object appears jagged on the generated image. This calculates the colors of multiple objects that exist in the subspace that are originally visible through that pixel as the color to be assigned to one pixel on the frame buffer (memory that stores pixels in one-to-one correspondence with the display screen). This is because, in order to simplify the color calculation, one point of one object in the subspace is used, although a value added according to the proportion of the area occupied in the subspace should be used.

FIG. 6 is an explanatory diagram of the occurrence of aliasing.

Assume that the actual space in which an image is to be generated is (a). Looking at the depth direction (z-axis direction), if the diagonal line area is in front of the white area, the white area behind the diagonal line area is hidden and cannot be seen.

On the other hand, as shown in (b), a total of 25 pixels (5×5) on the frame buffer are allocated.

In the hidden surface removal process, all pixels in the shaded area are colored black (actually, the color represented by the shaded areas) as shown in (C). The pixel data written as a result of this is an image in which the boundary between white and black is jagged (aliased) as shown in (d). The same applies when white is used as the color of all pixels in the white area.

However, originally, white and black parts should be added at a certain ratio, and the image should be schematically shown in (e). In other words, there should be gray pixels (a dotted pattern in the figure) at the boundary between white and black.

As methods for removing such aliasing, the following two methods are known.

■Supersampling method ■Filtering method (supersampling method) The supersampling method assumes a frame buffer that has finer pixels than the actual frame buffer as the color of one pixel on the frame buffer, and calculates the color for that pixel. The average value of the colors of a plurality of pixels corresponding to one pixel on the actual frame buffer is used.

FIG. 7 is an explanatory diagram of the supersampling method. (a
) is an enlarged view of the center pixel in FIG. 6(b). In other words, this figure shows one pixel. (b) assumes that pixels are virtually three times as fine in the vertical and horizontal directions as in (a).

When the color of this pixel is determined in the same manner as in FIG. 6, an image as shown in FIG. 6(c) is obtained. Here, if the value shown in (d) is used as the color value of each pixel, the average value is (OX3+9X6)+9-6, and this value is taken as the original value of (a).

(Filtering Method) The filtering method uses, as the color of one pixel on the frame buffer, the average value of the colors of a plurality of pixels in the vicinity thereof. FIG. 8 is an explanatory diagram of the filtering method. FIG. 6(a) assumes appropriate values for the colors of each pixel in FIG. 6(d). At this time, 1 from the top left to the bottom
As the color of pixel g located at the first pixel on the right,
As shown in (b), the average value of the colors of nine pixels in the vicinity of that pixel is determined.

The average value of the nine neighboring pixels is (OX5+9X4)+9-4. Since the nine pixels in the vicinity are taken, the calculation formula cannot be applied to the pixel at the very edge, so it cannot be calculated, but the nine pixels inside can be calculated in the same way. (C) shows the average value of pixels obtained in the same manner. The edges cannot be calculated, so they are not priced.

[Problems to be Solved by the Invention] The above-mentioned supersampling method has a problem in that increasing the pixel precision by n times increases the amount of calculations to n2, which is a high calculation cost. On the other hand, the filtering method
Although the calculation cost is not very high, the problem is that the entire image is blurred.

In the filtering method, in the image generation method using the Z-buffer hidden surface elimination method, as a method to prevent the entire image from blurring, the object being drawn is A method has been proposed in which the color of a pixel on the frame buffer corresponding to the contour is replaced with the average value of the color of neighboring pixels (Edge Inference).
with Application to A
ntialiasing, Computer Gras.
phics Vol, 17. No, 3.
T) p, 157-162. July, 1983)
. However, this method has disadvantages in that contour detection and tracking require high calculation costs and processing time is not constant.

The present invention has been made in view of these problems, and it is an object of the present invention to provide an aliasing removal method for image generation that can remove aliasing without blurring the entire image and without increasing calculation cost. .

[Means for Solving the Problems] FIG. 1 is a flowchart showing the principle of the system of the present invention. In the image generation method using the two-buffer hidden surface elimination method, the present invention provides the following steps: Referring to the contents of the Z buffer and frame buffer obtained as a result of image generation, in step 1), among the pixels on the frame buffer, the Z buffer is Apply a differential filter to detect regions corresponding to pixels with depth discontinuity (Step 2), apply an averaging filter to the detected regions (Step 3), and calculate the resulting average value of neighboring pixels. is replaced with a new pixel value (step 4).

[Operation] A differential filter is applied to the Z buffer to detect regions with discontinuity in the depth direction, and filtering is performed only on the detected regions. Since filtering is performed only on discontinuous areas, it is possible to provide an aliasing removal method for image generation that can remove aliasing without blurring the entire image and without increasing calculation cost.

[Example] Hereinafter, an example of the present invention will be described in detail with reference to the drawings.

Before explaining the example, we will explain image generation and hidden surface removal.

Explain aliasing. Image generation is a process of projecting the color of an object surface in a three-dimensional space, expressed by a model numerically stored in a computer, onto a two-dimensional image. At this time, depth information is lost. However, 3
When there are multiple objects on the line of sight in dimensional space,
If the object is not transparent, the rear (farthest from the viewpoint) surface cannot be seen, so only the front (closest to the viewpoint) surface needs to be projected.

Hidden surface elimination is a technique that prevents such invisible surface colors from being projected into an image. FIG. 2 is an explanatory diagram of the hidden surface elimination method. As shown in (a), when looking at a three-dimensional space from viewpoint 1, in the depth direction (Z-axis direction), Ml,
Assume that there are two objects M2. Here, 2. is the distance of Ml in the Z-axis direction, and z2 is the distance of M2 in the Z-axis direction.

When this is viewed from the front, as shown in (b), since Ml is in the front, a partial area (shaded area) of M2 that overlaps with Ml is not visible.

One of the hidden surface elimination methods is the Z-buffer hidden surface elimination method. In the Z-buffer hidden surface elimination method, the distance (depth) from the viewpoint is
Prepare a buffer to store , and set its contents to the maximum Z value that can be stored in the Z buffer.

Then, the color is projected onto the two-dimensional image for each object or surface, but before calculating the color of each point on the surface, the depth of that point and the corresponding pixel stored in the Z buffer are calculated. Compare depth.

If the depth of that point is greater than the contents of the Z buffer, the previously projected point on the surface is in front, so
Stop the color calculation, and if it is smaller, that point is currently closest to the viewpoint, so calculate the color, replace the corresponding pixel value on the frame buffer with that color, and replace the corresponding pixel value on the Z buffer. Rewrite the value of to the depth of that point. That is, the previously registered depth data is replaced with new depth data. As a result, the final image obtained is a projection of only the point on the surface that is visible in the foreground in three-dimensional space.

At the boundary where objects overlap, the two surfaces select the same pixel on the two-dimensional image as a projection target. Ultimately, the color of the surface close to the viewpoint is projected, and it was explained in FIG. 6 that this is the cause of aliasing. From this, it can be seen that aliasing occurs at the boundary of the object and that the discontinuous portion of depth stored in the obtained Z buffer corresponds to the boundary of the object.

FIG. 3 is a diagram showing the weights of the differential filter used in the present invention. The differential filter is composed of two filters: a vertical filter shown in (a) and a horizontal filter shown in (b). Each is weighted as shown in the figure. The image is scanned with these filters, and the values of pixels in the portion overlapping with the filters multiplied by the weighting are added together to obtain the values fx and fy of each pixel.

FIG. 4 is an explanatory diagram of filter calculation. It is assumed that the filter shown in FIG. 3(a) is superimposed on a pixel having a value as shown in the figure. The differential operation at this time is f x−−IXO+OX2+IX4+ (−1)xl+
OX3+IX8+ (-1)X5+OX2+lX4-1
0, which becomes the value of the middle pixel in FIG. When such calculations are repeated, the pixel 1 2 in the vertical direction becomes an emphasized pixel. Since this processing does not require random access to the frame buffer, images of the same size can always be processed in a constant amount of time.

It is clear that these filters output values close to 0 in areas where pixel values are constant, and output values with large absolute values in areas where pixel values change. That is, the absolute value of fx
Since the sum of the absolute value abs(fy) of (fx) and fy outputs a large absolute value in a portion where the depth information in the Z buffer changes, discontinuity in depth can be detected.

The averaging filter is a filter that functions as shown in FIG. 8, and is used for the purpose of suppressing sudden changes in pixel values in an image (frame buffer in this case). However, unnecessary averaging will only blur the entire image, so set an appropriate threshold for the differential value of depth obtained by applying a differential filter to the Z buffer, and if it exceeds a certain value, If the depth is less than 1, it is considered continuous,
The averaging filter is applied only to parts of the frame buffer where the depth is discontinuous. This makes it possible to remove only aliasing that occurs at the boundaries of objects.

Specifically, a mask image is generated in which the value of pixels whose depth differential value is greater than or equal to a threshold is set to 1, and the value of pixels whose depth is less than 1 is set to 0, and an averaging filter is applied to the image on the frame buffer. The mask image may be constantly moved, and the average value of the obtained color may be output in the portion where the mask image has 1 pixel, and the original pixel color may be output in the portion where the mask image has 0 pixel. This process can also be performed in a fixed amount of time similarly to the differential filter.

FIG. 5 is a diagram showing an example of a system configuration for implementing the method of the present invention. Image generation unit 11 using two-buffer hidden surface elimination method
The color information of the image generated in the Z buffer 12 in which depth information is stored is outputted to the frame buffer 13 and stored therein. The differential filter 14 performs differential processing on the depth information image stored in the Z buffer 12, and checks whether the differential value that is the result of the differentiation is larger than a preset threshold.

3 4 A mask image is generated in which the value of pixels with a larger differential value than the threshold is set to 1, and the value of smaller pixels is set to 0, and the mask image is stored in the mask buffer 15.
Output to. The averaging filter 16 simultaneously converts the image generated from the frame buffer 13 into the mask buffer 15.
Read the mask image from , calculate the average value of the pixels in the vicinity only for the pixels of the generated image that correspond to the pixels whose value is 1 in the task image, replace it with the average value, and output and write it to the frame buffer 17 . The contents of the frame buffer 17 are read out and displayed on the display section 18.

[Effects of the Invention] As described in detail above, according to the present invention, the boundaries of overlapping objects are subjected to differential processing on images with depth information, and the large value thereof is detected as a boundary area, and the detected area is By adopting a configuration that applies averaging processing only to
It is possible to provide an aliasing removal method for image generation that can remove aliasing without blurring the entire image and without increasing calculation cost.

[Brief explanation of the drawing]

Fig. 1 is a flowchart showing the principle of the method of the present invention, Fig. 2 is an explanatory diagram of the hidden surface elimination method, Fig. 3 is a diagram showing the weights of the differential filter, Fig. 4 is an explanatory diagram of the filter calculation, Fig. 5 6 is a diagram illustrating an example of a system configuration implementing the method of the present invention, FIG. 6 is a diagram illustrating occurrence of aliasing, FIG. 7 is a diagram illustrating a supersampling method, and FIG. 8 is a diagram illustrating a filtering method. In FIG. 5, 11 is an image generation section, 12 is a Z buffer, 13 is a frame buffer, 14 is a differential filter, 15 is a mask buffer, 16 is an averaging filter, 17 is a frame buffer, and 18 is a display section. 5 6 Flowchart illustrating the principle of the method of the present invention Figure 1 ■ OC (a) Pixel Φ value (b) Neighborhood (c) Explanatory diagram of average value filtering method

Claims (1)

[Claims] In an image generation method using the Z-buffer hidden surface elimination method, the contents of the Z-buffer and frame buffer obtained as a result of image generation are referred to (step 1), and among the pixels on the frame buffer, A differential filter is applied to the Z buffer to detect areas corresponding to pixels with depth discontinuity (step 2), an averaging filter is applied to the detected area (step 3), and the resulting neighborhood An aliasing removal method in image generation, characterized in that the average value of pixels in is replaced as a new pixel value (step 4).