WO1993008537A1

WO1993008537A1 - Process and circuitry for suppressing concealed picture elements

Info

Publication number: WO1993008537A1
Application number: PCT/DE1992/000881
Authority: WO
Inventors: Martin Cobernuss; Herbert RÜSSELER
Original assignee: Gesellschaft Für Mathematik Und Datenverarbeitung Mbh
Priority date: 1991-10-17
Filing date: 1992-10-16
Publication date: 1993-04-29
Also published as: DE4134576A1

Abstract

In raster graphics systems, a Z-memory is used and each Z-coordinate value is compared in a comparator (29) in order to suppress concealed picture elements. The disadvantage of this system is that the whole Z-memory (17) must be initialized or erased in order to ensure that only the valid Z-values of the actual image are compared. The required initialization phases take much time, reduce the available time for building the image and are especially disturbing in image sequences. In order to increase the image exchange speed, the Z-values entered in the Z-memory (17) are permanently allocated to the corresponding image by an image counter (23) and an image number memory (24). The Z-values are then only compared when the Z-values read out from the Z-memory (17) belong to the actual image. Otherwise, the generated Z-values are immediately entered in the Z-memory (17) and the corresponding picture elements are represented. According to the number of distinct images, the Z-memory (17) must be re-initialized or erased only when the available image number band width is exceeded.

Description

description

Method and circuit arrangement for suppressing hidden pixels

Technical field

The invention relates to a method and a circuit arrangement for carrying out processes that take place in a raster graphics system when comparing z-coordinate values to suppress hidden pixels.

State of the art

This is based on a prior art which can be found, for example, in US Pat. No. 4,924,415 or US Pat. No. 4,951,232 with regard to the basic functioning of raster graphics systems for generating and displaying images on video visual displays, which have a Z memory for storing the depth information - Z values - of three-dimensional objects. In such vision systems with computer support for the synthesis and display of two-dimensional images, such a Z memory is used to analyze object overlaps and to decide which object is in the background or in the foreground. This visibility method has also been known as the Z algorithm or Z buffer method for many years (see, for example: "Computer Graphics Principles and Practice", Foley et al., Addison-Wesley, 2nd ed. 1990, Pages 668 to 672).

In the development in this field, which has also been taking place for many years, it can be observed that many efforts are aimed at shortening the time by parallel, interlocking and / or forward-looking data treatment is necessary for the generation of individual images to be displayed.

Presentation of the invention

However, the invention is based on the problem of reducing waiting times which occur during a picture change and which are perceived as annoying especially in real-time animations and which can be attributed to the fact that the Z memory has so far been deleted and reinitialized with each picture change sieren is. It should also be borne in mind that when the resolution of a vision system is increased, the storage capacities and thus the deletion and initialization times, that is to say the waiting times for image changes, increase disproportionately.

This problem is solved in a method of the type mentioned at the outset according to the invention by means of the measures listed below for increasing the image change rate of image sequences: b) assigning the current image number when storing z-coordinates to be displayed; c) comparing the image number of a stored z coordinate value with the image number of the current image for

Pixels of the same x, y coordinates when generating the data of a current image to be forwarded to a vision system, and: dl) if the compared image numbers match: performing the comparison of the z-coordinate values in order to suppress a hidden pixel, optionally suppress the hidden pixel and storing the zS coordinate with the assigned image number of the pixel to be displayed and forwarding the data of the pixel to be displayed to the vision system, o the: d2) if the compared image numbers differ: storing the z coordinate with the assigned image number of the current image point and forwarding the data of the current image point to the vision system without prior comparison of z coordinate values; e) Deleting or overwriting of stored picture numbers which would be repeated if the available picture number bandwidth Fmax = (2exp n) - 1, with n = bit depth, were repeated.

A circuit arrangement for this, in which the equipment has a control unit and at least one Z memory and a Z value comparator, is constructed in accordance with the invention in such a way that an image counter, the counter reading of which is included with every image change, an image number memory and an Image number comparators are available which are connected such that the x, y coordinates of the current image points, as well as the Z memory, are present at the address inputs of the image number memory and the output of the image counter is connected to the data inputs so that the output of the image number memory is connected to an input of the image number comparator and the output of the image counter also leads to the other input of the image number comparator, and that the already existing control unit for the control of processes in the Z memory and Z value comparator also for the control of the frame counter and the write and delete operations of the frame number memory s designed and equipped with inputs to which the signals of both comparators and an image change signal are routed.

The main idea is the following: The mode of operation of the Z algorithm is based on a comparison operation of all generated Z values with the corresponding Z values from the Z memory. The Z values read out must at be relevant to the image in question. The deletion phases customary up to now therefore delete all entries from the previous image and thus ensure correct comparison values. However, this effect can also be achieved with the help of a clear assignment of the Z values entered in the Z memory to the respective image. During image generation, the Z values of the new image points are only compared with the read Z values if they belong to the current image. In the other case, the generated Z value can be entered immediately in the Z memory and the pixel can be displayed, since no pixel that belongs to the current image has yet been generated at this position. In the conventional Z comparison, a comparison operation always takes place and the Z memory must therefore always be deleted after each picture, so that the new picture is not influenced by invalid Z values. The technical teaching of the invention ensures that all Z values of an image can be identified, global deletion of the Z memory is therefore no longer necessary and the deletion and initialization times decrease.

In the circuit arrangement according to the invention, the pixels can be assigned to the respective image with the aid of an image counter and an image number memory which records the current image number for each pixel. If there a counter reading is assigned to the current picture, which the picture number memory does not contain in relation to the non-current pictures, then all picture points which do not belong to the current picture can be clearly identified and excluded from a Z comparison operation.

Advantageous embodiments of the invention are specified in the subclaims. The solution according to the invention of being able to increase the frame rate of frame sequences by reducing the deletion time of a Z memory can be Men who work with a programmed Z algorithm, or use in systems with specific circuit arrangements.

Brief description of the drawings

An embodiment of the invention and conventional general prior art are shown in the drawings and are explained in more detail below. First, the exemplary embodiment of the invention is described.

Show it:

Figure 1: the flow diagram of the method according to the invention; Figure 2 is a block diagram of a raster graphics system, with integration of the Z memory in the vision system; Figure 3 is a block diagram of a hardware implementation of the

Invention; Figure 4: a 2D representation of several overlapping objects, and

Figure 5: the flow diagram of a conventional Z algorithm.

Ways of Carrying Out the Invention

The method according to the invention runs as follows with an image counter and an image number memory in cooperation with a conventional Z memory (FIG. 1): after switching on, the image number memory is initialized with the initial value of the image counter, for example "0" (Step 6). The Z memory is not initialized or deleted! The frame counter is then incremented (step 7). The image counter reading indicates the sequential number of the image currently being generated. These operations must be carried out once after each switch on or reset. The genes- The image points of an image can now begin. The end of image generation and the beginning of a new image is indicated by a central signal or by an additional marker at each pixel. As long as no image change occurs, the image counter reading is read from the image number memory for each pixel generated and compared with the current image counter reading (step 8). If both image numbers are the same, there is already a pixel in this position that belongs to the current image, and a Z comparison must be carried out (step 9). Depending on the Z comparison, the old Z value is replaced by the new Z value and the image point is displayed (step 10), or the old Z value and thus also the old image point are retained. If the picture counter reading from the picture number memory is different from the current picture counter reading, there is still no picture point of the current picture at this position and consequently no Z comparison is necessary. The Z value of the new pixel is written into the Z memory and the pixel is displayed (step 10). At the same time, the current image counter status is entered in the image number memory for this pixel (step 11). This shows that this position is occupied by a pixel of the current image. All pixels are processed in this way. These processes function as long as only counter readings that are different from the current picture counter reading are noted in the picture number memory, based on the non-current pictures. With an n-bit image counter and an n-bit wide image memory, Fmax = (2 exp n) -l images of an image sequence can be distinguished. Image "0" with the image counter reading of, for example, also "0" corresponds to the initialization state after switching on. When writing the first image, the image counter reading ("1") is now compared with the content of the image number memory ("0" or "1"). During the second image - image toughness Status "2" - with the content of the frame number memory from "0", "1" or "2" and so on. The correct execution of the algorithm is ensured if there are only values in the picture number memory that differ from the current picture counter status.

After Fmax pictures the picture counter status would be "0" again and there could still be entries in the picture number memory with the picture counter status "0". There are two possibilities to ensure that the frame counter is always different from the possible entries in the frame number memory:

Case -a): According to Fmax Bild, the entire image number memory is again with the initial value of the image counter e.g. Initialized "0" and the frame count increased by "+1" (case a, step 12, corresponds to the initial initialization). In practice, however, this means that the memory bandwidth of all Fmax image cycles is considerably reduced. In complex scenes, this can result in the image sequence being periodically "braked" with the period Fmax and then "accelerating", and this is regarded as an undesirable effect.

Case - - b): In order to distribute the memory bandwidth as evenly as possible, it is advantageous to divide the deletion process. After each picture, 1 / Fmax of the entire picture number memory is continuously written with the current picture counter status (case b, step 13) and only then is the picture counter modulo Fmax +1 incremented (step 14).

If the frame counter length is specified with n, the counter runs from 0 to Fmax, with Fmax = (2 exp n) -1. Before each new image, the 1 / Fmax part of the image number memory is written continuously. The time it takes to delete per picture is reduced to 1 / Fmax of the complete deletion time.

The raster graphics system with real time properties shown as a block diagram in FIG. 2 is suitable for successfully implementing the solution according to the invention. The visualization processes necessary for image generation are supported by conventional specific hardware. The visualization is carried out by a pipeline, consisting of a geometry processor 15 (affine and perspective transformations, backfacing, 3D clip process), a renderer 16 (calculation of all pixels of a surface, surface shading according to Phong), the Z Memory 17 (elimination of the hidden pixels), and from a 2D subsystem 18 including image repetition memory and display device 22.

In the following, only the Z memory 17 and the components necessary for reducing the deletion time are dealt with.

The Z memory 17 is an independent unit within the visualization pipeline, between the scan converter - renderer 16 - and the image repetition memory 18. The Z memory 17 evaluates the Z values 19 of the pixels generated and conducts them if they are successful Z comparison the x, y coordinates 20 as addresses. and the color value 21 of the pixels to the image repetition memory 18.

3 shows, for the purposes of the invention, an image counter 23, an image number memory 24 and an image number comparator 25, for example for n = 4, Fmax = 15 tern. A control unit 26 carries out each Z comparison in accordance with the sequence shown in FIG. 1 (case b). However, the sequence in embodiments of the invention is summarized that BZ. old from the picture number memory 24 and Z. old from the Z memory 17 can be read simultaneously. Since a picture number is compared to 4-bit in picture number comparator 25 significantly faster than the 24-bit Z comparison in Z-value comparator 29, in many cases the result of the 24-bit Z comparison does not have to be be waited for. Furthermore, the shallower memory of frame number memory 24 allows the use of fast static RAMs, so that the average time for a memory cycle including the Z comparison is considerably reduced.

After the generation of an image, the 1 / Fmax part of the image number memory 24 is written by the image change signal 27 with the current image counter reading. The control unit 26 generates the delete addresses 28 required for this for the picture number memory 24. The picture counter 23 is then incremented and the generation of the new picture can begin. The proportion of the deletion time in the total time for the generation of an image is only approx. 1.5% if 10 images are generated per second; This means that approximately 98.5% of the time is available for the Z comparison.

This advantage of the solution according to the invention is particularly effective in real-time animation systems and the like. In conventional systems, considerable problems arise in this regard, as can be seen from the following explanations in connection with the description of FIGS. 4 and 5.

In the example according to FIG. 4, three objects are made visible on a monitor, which partially hide one another - object 2 in front of object 1, or penetrate one another - object 1 and object 3. The Z memory contains the for each pixel Depth information (Z coordinate or Z value), with the aid of which a decision is made as to whether a generated pixel is visible or hidden by the pixel of another object. This decision is made by comparing the Z values of all new pixels Z.neu with the respective corresponding Z values, the pixels already generated from the Z memory Z.alt. Depending on the selected origin of the coordinate system, Z.neu greater than Z.alt or Z.neu smaller than Z.alt (in FIG. 4, Z = 0 in the eye of the beholder) covers the old pixels with the new pixels , as well as the transfer of the new Z values into the Z memory. Otherwise there is no exchange of the Z values and the new pixels are not displayed.

However, the comparison of the new Z values with the old Z values from the Z memory is only correct if the old Z value read out is valid. In two cases, however, this is not the case: a) after switching on and b) when generating a new image. After switching on, the Z memory contains only random values, so that a correct comparison with the Z values of the pixels of the first image is not possible. The entire Z memory must therefore be initialized. A customary method for initializing the Z memory consists in writing an initialization value which does not belong to the range of values of the Z values and which always leads to an exchange. Z.min or Z.max is usually selected depending on the coordinate origin.

In the representation of the flow diagram of a conventional Z algorithm in FIG. 5, the entire Z memory is initialized with Z.max after switching on (Init) in step 4. As a result, each newly generated pixel lies against this background and is exchanged. In the conventional Z algorithm, however, the Z memory must also be reinitialized or deleted before each new image in order to have all the Z values of the remove the old image so that it does not affect the new image (step 5).

The initialization phases represent a not inconsiderable portion of the total time for the generation of an image, in particular for viewing devices with high resolution with the necessary size of the Z memory of approx. 4 Mbyte (1280 x 1024 pixels a 24 bit) Real-time animation systems reduce the initialization phases and the time available for building an image. This reduces the complexity of the images that can be represented and the overall system performance is reduced. A reduction or saving in the initialization phases to which the invention relates therefore makes a not insignificant contribution to the increase in the overall system performance.

Other conceivable measures are less effective. For example, Z memory systems based on dynamic RAMs (cycle time approx. 120 ns) require approx. 15 ms to 20 ms for the initialization. In real-time animation systems with image sequences of at least 10 Hz, the available image construction time is reduced from 100 ms to 8C ms to 85 ms, which corresponds to 15% to 20% of the total time.

Video RAMs represent an alternative. These are dynamic RAMs in which, among other things, an entire row of the internal memory matrix of, for example, 256 memory cells can be written in one memory cycle. If the Z memory is built up with video RAMs, the deletion times are 20 times less than those of the method according to the invention. However, another fact has a negative effect: Video RAMs currently have cycle times of 180 ns to 200 ns when accessing memory compared to 110 ns to 130 ns of normal dynamic RAMs. Based on a memory bank, the maximum memory bandwidth is 5.5 M image points / s Video RAMs at 9 M image points / s with dynamic RAMs (READ or WRITE, not RMWl). This corresponds to a performance difference of 50% to 60%, so that video RAMs are not a serious alternative to dRAMs, despite their favorable deletion times.

Claims

1. A method for carrying out processes that take place in a raster graphics system when comparing z-coordinate values to suppress hidden pixels, the measures listed below for increasing the image change rate of image sequences: a) numbering the individual images of the image sequence; b) assigning the current image number when storing z-coordinates to be displayed; c) comparing the image number of a stored z coordinate value with the image number of the current image for pixels of the same x, y coordinates when generating the data of a current image to be passed on to a vision system, and: dl) if the compared ones match Image numbers: performing the comparison of the z-coordinate values for suppressing a hidden image point, possibly suppressing the hidden image point and storing the z-coordinate with the assigned image number of the image point to be displayed and forwarding the data of the image point to be displayed to the vision system, or: d2 ) If the compared image numbers deviate: storing the z coordinate with the assigned image number of the current image point and forwarding the data of the current image point to the vision system without prior comparison of z coordinate values; e) Deletion or overwriting of stored picture numbers which would be repeated if the available picture number bandwidth Fmax = (2exp n) - 1, with n = bit depth, were repeated.

2. The method according to claim 1, characterized in that after an image sequence of Fmax images all stored pixel-related picture numbers are deleted and the storage medium for the picture numbers is reinitialized.

3. The method of claim 1, d a d u r c h g e k e n n z e i c h n e t that continuously at each image change at least 1 / Fmax of the total stored pixel-related image numbers are replaced by the current image number.

4. Circuit arrangement for carrying out processes which take place in a raster graphics system when comparing z coordinate values to suppress hidden pixels, and in which the equipment comprises a control unit (26) and at least one Z memory (17) and one Z -Value comparator (29), characterized by the fact that an image counter (23), the count of which is incremented with each image change, an image number memory (24) and an image number comparator (25) are present, which are switched in such a way that the image number memory (24) has the x, y coordinates of the current pixels at its address inputs, as well as the Z memory (17), and the output of the image counter (23) is connected to the data inputs so that the output of the Frame number memory (24) is connected to an input of the frame number comparator (25) and the output of the frame counter (23) is also connected to the other input of the frame number comparator ( 25) leads, and that the already existing control unit (26) for the control of processes in the Z memory (17) and the Z value comparator (29) also for the control of the image counter (23) and the write and Deletion processes of the picture number memory (24) are designed and for this purpose equipped with inputs to which the signals of both comparators (25, 29) and a picture change signal (27) are led.