JP3089356B2 - Beam forming device for matrix display device - Google Patents

Description

Description: FIELD OF THE INVENTION The present invention relates generally to the field of computer generated image generation, and more particularly to a method for preventing aliasing of computer generated images. In general, the application of an anti-aliasing filter to the output of a partial pixel image memory configuration is described in order to improve the image quality of a display device, in particular a color mosaic display device. As a result, a smoother and more stroke-shaped image is obtained.

[Problems to be solved by conventional technology and invention]

A problem that occurs with flat panel matrix displays and / or raster displays involves the aliasing shown in the simulated display 11 of FIG. 1 due to the discontinuous nature of some display systems. This results in an output that is discontinuous, ie, jagged, as in line 10. This is in contrast to the widely accepted stroke 12 of high quality, where each line is smooth and continuous (ie, preventing aliasing), as seen on CRT displays. Unless aliasing is prevented, computer generated images on raster displays will exhibit disturbing artifacts such as moiré patterns, rainbow effects, jagged edges, discontinuous motion, missing details and / or scintillation. Normal. These artefacts are all
It also highlights a more general problem called aliasing. Aliasing occurs when the details of the image to be displayed, the spatial resolution content, exceed the sampling bandwidth capabilities of the display system. Fortunately, sampling theory makes it possible to eliminate aliasing altogether. In order to eliminate aliasing, the signal to be sampled must first be band-limited so as not to include frequencies exceeding 1/2 of the sampling frequency of the system. To prevent artifacts from appearing in the image, the image must first be band limited so that the spatial frequency content of the image itself meets this criterion.

In a computer generated image system, an image can be represented as a signal whose amplitude varies as a function of position along the surface of the display. As the image becomes more detailed, this signal changes more rapidly along the screen. Increasing the rate of rise and fall of the signal has the effect of pushing the spectral energy content of the image to a higher spatial frequency range. Therefore,
Band limiting the computer generated image means that the computer must be programmed to prevent abrupt transitions. For example, pixels near a polygonal line, ie, a detail such as a side, are shaded so that the signal moves up and down at an appropriate rate. The appropriate speed is determined using sampling theory, as described above.

2. Description of the Related Art Conventionally, an attempt has been made to mathematically determine an appropriate intensity for each pixel of a display device in a CPU (central processing unit). Weighted filter for window area /
The convolution was performed and the result was used as an indicator of the appropriate strength value. Some hardware is implemented by applying an intensity value to pixels on both sides of the desired pixel. This method is usually selected when using a vector generator to render graphical symbols.

The most commonly used in the prior art is the write cycle architecture. That is, while the image is being written to the memory, the computer band limits the image. Over the years, many different algorithms have been developed to implement anti-aliasing. One of the common methods among them was based on how far a pixel was from a given feature. A characteristic, for example, a distance from a side is determined for each nearby pixel. The further away a pixel is from a side, the less the side affects the intensity assigned to that pixel. As a result, the luminance transition was smoothed, the spectral content of the image was appropriately restricted to lower frequencies, and aliasing artifacts were eliminated.

According to another well-known method, which is theoretically more precise than the above, assigns intensity or color values to adjacent pixels as a function of the area of overlap. That is, the larger the portion of a pixel covered by a particular feature, the greater the effect the feature will have on that pixel. For example, a pixel that is completely covered by a polygon is assigned 100% of the value required by that polygon. If, on the other hand, the pixel is on a side of the polygon and only covers one third of the pixel, the pixel is illuminated with only one third of the value specified by the polygon. Will be done. This method can also perform an effective reduction filter operation on the image, limiting the spatial frequency content as needed. In addition, anti-aliasing was similarly performed appropriately.

As a further alternative, a filter convolution method, which is most accurate for non-real-time applications, was also employed. This reduces the bandwidth of the image by sliding the windowing kernel along the source image with a higher resolution than the desired output resolution. Usually, 3 × 3 processing nuclei were selected. The processing kernel stores a number corresponding to a specific point spread function. The transfer function and the shape of the point spread function are derived from sampling theory. By sliding the nucleus pixel by pixel along the source image,
It runs the convolution process directly. This convolution of the nucleus using a high resolution source image results in an output image having a spatial bandwidth that matches the spatial bandwidth of the sampling display system. Therefore, aliasing was also avoided in this case.

All three of the above anti-aliasing methods are based on write cycles. This creates significant deficiencies when it is necessary to provide high resolution real-time performance. An important issue is its speed. When anti-aliasing is performed, even if the resolution is not so high, for example, a 512 × 512 system, the burden on the system is large, and the write cycle architecture is pressed to the limit. The processing elements must be chosen at a much higher cost for much higher performance. Of course, you can use a parallel approach to alleviate some of the issues,
In that case, the flexibility of the control and the system becomes a big problem.

When color mosaic display technology is used in the display device, the architectural problem of the write cycle becomes even more important. The resolution of this generation method is considered to be several times higher than the resolution that has already been found to be difficult, and there is a tendency for higher resolution. In addition, the particular color pattern of the mosaic must be taken into account by the graphic generator in order to be able to successfully render the antialiased image. Considering the color pattern allows the image generator to adapt the frequency content of the image to the specific capabilities of a given color mosaic display. To summarize the background, the prevention of aliasing based on write cycles can be used to generate high-resolution real-time images, especially for color mosaic displays.
The conclusion is that this will result in considerable cost.

[Means for solving the problem]

The present invention performs effective anti-aliasing by approaching the problem from a different perspective than the prior art described above, that is, from a system level perspective. In particular, rather than burdening the computer with the extremely repetitive task of filtering the image as image details appear, the present invention provides:
A new element 23 on the output side of the image memory, ie in the reading side in a commonly known way (FIG. 3)
As many redundant activities as possible. In the present invention, each active pixel in the image memory is extended to a Gaussian profile in the beamformer 23 (ie, a new element). This is represented as 13 in FIG. 2 which shows the extension pictorially.

The new system element is called a beamformer. This device undertakes the extra burden of performing anti-aliasing and operates as a matching filter between the graphics generator and the display 22. The beamformer limits the bandwidth of the image according to the sampling theory, and uses the partial pixel information as a means to extend the apparent sampling frequency of the system, especially with respect to the display head when using a color mosaic display. In this new system,
The beamformer is between the display and the computer (FIG. 3b). The important features of the beam forming apparatus are: a) the impulse data from the image memory
Extending to a point spread function 13 with a bandwidth within the capabilities of the sampling system that can be configured as 2, b)
Using the partial image information obtained from the graphic generator 24, the center of gravity of the brightness of the point spread function 13 is adjusted so as to appear between sampling points that can be considered as sampling points of the display device 22, Thereby, a higher position addressability is provided.

Inside the computer, only a few functional modules need to be modified to make this new system approach work. Referring to FIG. 3b, the first functional module, the graphic generator 20a, must provide partial pixel information in addition to the standard integer location address output. Partial pixel information is usually available in the prior art, but is often not used outside the graphics generator.

The other module, image memory 21a, must be modified to include a storage device that stores the new partial pixel information provided by the graphics generator. The image memory must store the partial pixel information in addition to the normal color information and attribute information. Therefore, the present invention provides a beam forming apparatus as a new element in addition to these deformed elements 20a and 21a of the computer.
23 will be provided.

The advantages of the present invention are that the processing throughput is significantly improved, the system is simple, and the costs are reduced, especially with regard to color mosaic display technology.
Yet another advantage of the present invention is that it most effectively prevents aliasing in color mosaic displays. Furthermore, the present invention allows the graphics processor to be independent of the resolution and pixel arrangement of the display device.

That is, the beam forming device 23 operates on the data obtained from the image memory, and generates data to be displayed on the display device, which is correctly prevented from being aliased. The beamformer looks at one window of data from the image memory and calculates the appropriate intensity value to be displayed. Also, the anti-aliasing function is moved from the graphics generator to a separate dedicated hardware part. Beamformers solve some of the problems of conventional systems. Since the shape generator does not need to perform any anti-aliasing work,
The throughput of the system is improved. Also, the beamformer performs all of the processing associated with the color mosaic display, thereby reducing the workload of the graphics generator, thereby further improving the throughput of the system.

〔Example〕

FIG. 3A shows a graphic generator 20 and an image memory 21.
And a conventional display system having a matrix display device 22. FIG. 3B shows a preferred embodiment of the present invention. It includes a deformed graphic generator 20a, a deformed image memory 21a, and a filter element 23 of a beamformer, along with a display device 22. FIG. 5 shows the deformed figure generator 20a. In addition to providing standard integer address values 50 for x and y, thereby addressing the image memory, the modified graphics generator further provides partial pixel data, ie, partial address data 51.
The partial pixel information, that is, the partial address information 51, can be used to instruct the subsequent system element, the beam forming device 23, to adjust the light energy to a position between the individual pixels. With respect to sampling theory, partial pixel information provides additional information about the sample points to be written to the image memory. This indicates the phase of the sampled amplitude data. Referring to FIG. 4, the partial pixel information can be used to guide light energy to any one of the 16 positions in one pixel 31 of the image memory 32, for example. An enlarged view 30 of pixel 31 shows a partial pixel that can be made available to the display system to more precisely position the light energy forming the image. For example, partial pixels
33 may be indicated by a graphic generator and stored in a modified image memory.

Next, FIG. 5A will be described. FIG. 5A shows a modified form of the image memory 21a. Instead of storing only the color data or the attribute data 71, the modified image memory further stores the partial pixel address information 72 obtained from the modified graphic generator. Later, at the time of readout, the image memory supplies standard video information along with partial pixel information to the beamformer.

Referring again to FIG. 2 or FIG. 3b, the beam forming device 23 is shown. The relationship between the input and output of this device is
2 shows that the device performs an operation very similar to a standard two-dimensional image processing filter. The impulse applied from the image memory to the input terminal of the beamformer originates in an expanded manner from the output terminal of the beamformer. This extension is the point spread function of the beamformer corresponding to the spatial frequency transfer characteristics of the two-dimensional image processing filter described above. In other words, each active pixel in the image memory 21a is expanded in the beam forming device 23 into a Gaussian profile. Data coming from the image memory 21a is collected by the beamformer 23 and interpreted to determine the appropriate intensity level for each pixel.

There are at least four differences between a beamformer and a conventional two-dimensional pixel processing filter: The beamformer extends two-dimensional image processing filter technology to: 1) generally prevent aliasing of raster images; Solving the problem; 2) in particular, solving the problem of anti-aliasing of the color mosaic display; 3) the beamformer uses a point spread function extracted for a given pixel pattern of the color mosaic; The center of gravity of the diffusion function can be finely positioned under the instruction of the input partial pixel information. (That is,
The filter coefficients are a function of the input data, as opposed to a standard filter implementation. Another important feature of the present invention comprising a deformed figure generator, a deformed image memory, and a beam forming apparatus is its scan conversion capability. Since the beam forming device 23 constitutes an interface between the computer and the display device, the drawing elements of the computer, that is, the graphic generator 20a, and the image memory 21a must be configured independently of the resolution and pixel arrangement of the display device. Can be. There is no need for a one-to-one correspondence between storage locations in the image memory and display pixels. The beamformer performs an interpolation function between the computer and any raster display technology. Since this is an evolving display technology, it is particularly useful in the case of a color mosaic display device in which the resolution and pixel arrangement may change from one display head to another. In that case, the system will have the advantage that the computer will be stabilized and not subject to such developments and changes. Another secondary advantage is that graphics processing can be performed at a lower resolution, which can achieve even greater improvements in the effective throughput gained by releasing only the burden of anti-aliasing. That is the point. further,
The anti-aliasing data is added at the output of the image memory and, therefore, cannot erase the anti-aliasing information below it, ie, previously drawn / written. No read / modify / write cycle is required. As a result, the effective processing throughput is further improved.

The beamformer uses the partial pixel positioning information from the image memory to more precisely position light energy relative to the display. The beamformer can more precisely position the light energy by adjusting the centroid of the brightness of its output point spread function. Since the point expansion function corresponds to a collection of pixels, the amount of energy from the pixels can be mutually adjusted between the pixels, and in doing so, whether the energy is generated between the pixels It is possible to look like. In a preferred embodiment, the beamformer positions the light energy within one third of the visual angle. This is usually related to one quarter of the screen width / height for positioning accuracy. The device shows the vernier adjustment capability available to eliminate sampling artifacts of the display system.

Using a beamformer also eliminates the need for the graphic generator to know the color arrangement of the color mosaic display, as in the case of a color matrix liquid crystal display (LCD). Each pixel in the image memory can be processed independently of the pixel arrangement in the LCD. Colors can be written in the image memory regardless of the pixel arrangement of the display device. This was previously possible, but was limited to non-color mosaic display technologies such as CRT display technology. The beamformer may incorporate a color sequencer and a lookup mechanism to separate the colors from the image memory into the components of the primary colors that make up the color mosaic pattern of a particular display. In this way, different pixel patterns can be easily managed, so that the color mosaic display technology can be processed as easily as the CRT technology. Typical primary color patterns used in color mosaic displays include vertical RGB stripes, diagonal RGB stripes, delta or hexagonal RGB, quad RGBG, Honeywell quad RGBY and quad RGBW. RGB is red, green,
Represents blue, Y indicates yellow and W indicates white.

The beam forming device 23 defines an anti-aliasing strength level for each primary color. For example, in a raster CRT display, the intensity levels for the red, green and blue channels would each be interpreted independently of the other two color channels. Thus, color continuity is obtained through the use of a beamformer.

FIG. 4 shows an enlarged view 30 of one pixel 31 in the image memory 32. In this case, the ability to address partial pixels can be used to improve the apparent resolution. FIG. 4 shows, by way of example, 16 partial pixels for each pixel. When the partial pixels can be addressed, it is possible to finely position the luminance profile as indicated by 33 in the expanded pixels 30. Thus, the shape and position of the line will be more accurate. The partial pixel information is stored in the image memory 21a, and the beamformer 23 uses the information to precisely position the Gaussian beam center. This increases the effective resolution of the display device 22 as the number of bits of partial pixel addressability increases. A 512 × 512 image memory with 3-bit x and y addressability can locate one pixel with the effective resolution of a 4K × 4K image memory.

The beamformer 23 forms a new system using a fixed configuration in the graphic generator 20a and the image memory 21a, and only needs to be modified in the beamformer to drive various different display devices. For example, when transforming the LCD display device from the RGB diagonal array to the RGBG array, it is only necessary to change the color coding method of the beam forming device 23, and the graphic generator 20a will not need to be modified. The graphic generator renders an image with the concept that each pixel may be of any color. Although the color arrangement of the pixels is determined by the beamformer 23, in the preferred embodiment the beamformer reads the pixel arrangement from a small ROM. Different sized displays can be driven by the same graphics generator system without having to modify hardware or software. With the change of the display device, only the beam forming device is modified.

A beamformer is a device that is located between an image memory and a raster display device and is designed to perform anti-aliasing on the display device. The beamformer includes a memory for providing a window of data. The theory of operation underlying the beamformer is that in the image memory, a window of data centered on the next output pixel is examined and the appropriate output pixel is determined based on the data inside the sliding window. It is to determine the strength level.

Next, FIG. 6 will be described. FIG. 6 shows an example of a sliding window used for processing in the beam forming apparatus, and the resulting pixels displayed on the apparatus. This figure shows an example of the operation of the beam forming apparatus. The beamformer retrieves a window of data from the image memory and performs the appropriate processing to generate a single output pixel. The sliding window is moved line by line from left to right and from top to bottom, as in a normal raster display.

The beamformer typically includes several line memories to perform access to a window of data centered on the next output pixel. The size of the window is determined by the geometric shape of the pixel pattern of the display and the desired line width of the display. In the example shown here, a 6 × 6 window is used.

In the first step of the operation of the beamformer, the active elements in the window are classified by color. FIG. 7 shows a block diagram of the components of the beam forming apparatus. The beam forming device 23 includes an element 50 for selecting the intensity from the color bits.
, A shaded up element 51, a maximum shade determining element 52, and an element 53 for moving the sliding window. First, a color process is performed to classify data according to the next output pixel (only for a color mosaic display device). Second, for each active pixel in the window, perform a shaded lookup based on a Gaussian profile. Third, all shades are compared and the maximum shade is determined. This maximum shade is the output to the display. FIG. 8 shows an example of the contents in the slide window before color processing. The slide window overlaps an image in the image memory that includes a yellow (green and red) line and a red line. FIG. 9 shows an example of the contents of the slide window after color processing for the next output pixel of green, and FIG. 10 shows an example of the slide window after color processing of the next output pixel of red. Is shown. Raster where each pixel has one color
In the case of a CRT, the processing is red on the first channel,
Are executed in parallel, such as blue on the third channel and green on the third channel.

After performing the color classification, the intensity is determined for each active pixel in the sliding window using a look-up table method that determines the intensity based on the distance from the center of the window. To determine the intensity for each active pixel, use a Gaussian profile compatible with the display.

FIG. 11 shows an example of a Gaussian profile inside a sliding window. Profile ring 60, 61, 6
2,63,64,65 and 66 each represent one intensity level, just as a topographic map represents elevation data by rings. For each active pixel in the sliding window,
The strength level is calculated based on the Gaussian profile. When determining the intensity level, the partial pixel data is used. Using the partial pixel data, the intensity is looked up in the main block shown in FIG. The next step in the process is to determine the maximum intensity level from all active pixels. This is performed using a comparator or another look-up table. Use the maximum intensity level as the output intensity level.

12 to 17 can be considered as a group of time frames. FIG. 12 shows one line of the sliding window 72.
One data point 83 of the image memory during filtering on 83 is shown. All other points are not relevant to the image. The thick lines 80-86 and 90-96 are in this case 6 ×
6 represents pixel boundaries in a six pixel window. 87,9
Thin lines such as 7 represent 16 partial pixel addresses for each pixel in this example of FIG. Circles 60-66 represent the limits of each strength level. In FIGS. 12 to 17, the window 72 slides to the right by one pixel for each successive frame, so that as one group of frames elapses, a single data point 70 apparently appears. It will move to the left. The following table shows the six consecutive frames of FIGS. 12 to 17 and the intensity and output intensity contributed by this single data point 70.

The intensity value to be used is determined using the partial pixel address. Strength is based on radial position, just like a topographic map. The innermost circle represents a strength level 7 and goes down one level per contour ring to 0 outside the outer circle.

FIGS. 12 to 17 show examples of a single data point in the image memory when being filtered on one line of the sliding window. For reference, there is shown one example of a plurality of adjacent data points in an image memory as filtered through a sliding window during a single pass along the screen. These nine figures represent sliding windows for nine consecutive time frames. First, perform a strength check for each point, then
Perform a comparison to determine the maximum strength. In this case, the maximum intensity value determined from comparing all the active points in the window is used as a value output to the display device. Table 2 below summarizes the three points 97, 98 and 99 of the image memory that are close to each other when filtered through a sliding window on a single line.

For lines of different luminance profiles, anti-aliasing data can be generated using the beamformer 23. This is a Gaussian profile change. Multiple brightness profiles can be used with one beamforming device. The processing of multiple brightness levels differs only in the initial Gaussian profile used to determine the intensity level for each active pixel. A look-up is performed for each pixel based on the position of the partial pixel and the setting of the intensity level. Next, all intensity levels are compared to determine the maximum intensity level to be used as output. Different brightness levels are processed in a look-up table.
FIGS. 27, 28 and 29 show three different brightness intensity profiles. In order to prevent aliasing of multiple brightness levels, it is usually considered that many such profiles are incorporated into the beamformer.

Modified Vector Generator In the present invention, the vector generator is a dedicated graphic generator. Digital vector generators typically render the lines and graphics by calculating the addresses of the pixels in an incremental manner and then commanding each of those pixels to perform a write operation. Solve the general formula (y1-y0) = m (x1-x0) and use it as the basis for calculating the address for each image. This equation is decomposed into equivalent parameter equations x _i = x _o + d _x * i and y _i = y _o + d _y * i. Note that m = d _y / d _x ,
Normally, either d _y or d _x is set to a value equal to one.

FIG. 30 shows a conventional image memory (0,0) to (7,
Show the vector drawn up to 5). The vector generator solves the basic equation to find the value of m. Since Delta Y is less than Delta X, this term is used to increment the Y address as the vector generator steps along the line. The X address is incremented by one each time. (If Delta Y> Delta X, Y is incremented by 1 and X is incremented by 1 / m.) FIG. 31 shows a functional block diagram of a typical vector generator for the X address. First, the starting point is loaded into the accumulator. At each clock cycle, the accumulator adds the previous accumulator output (stored in the latch) with an increment specified by delta. X of next pixel
To generate an address, the output of the latch is provided to a rounding circuit. The accumulator has sufficient precision to handle fractional parts. The rounding circuit rounds the number to the integer required by the pixel addressing circuit (eg, 2
Round 7.862 to 28. The same circuit is used for generating the Y address. In the case of FIG. 30, the initial values and increments for X and Y are as follows.

Initial Value Increment X 0,0 1,0 Y 0,0 0,714 Using a vector generator with a beamformer,
The decimal information available from the accumulator can be used. FIG. 32 is a block diagram of a vector generator used with the beamformer. This is similar to that shown in FIG. 31, but with two differences. Instead of the rounding circuit shown in FIG. 31, a truncation circuit is used to generate the pixel address. The latch output is provided to a new block that removes the fractional part of the pixel address. This decimal address information is stored in the pixel as partial pixel data together with a data field of normal color and intensity. Partial pixel data is accumulated in each pixel. For example, 27, 86
2 is rounded down to 27 as an integer address, and 0 and 862 are stored in the pixel as partial pixel data.

FIG. 33 shows the vector rendered using the deformation vector generator. By accumulating partial pixel data in each pixel, each pixel can be positioned very accurately.
FIG. 33 shows 16 partial pixel positions for each pixel. No. 33
Comparing FIG. 30 with FIG. 30, it can be seen that the partial pixel data positions the line center more precisely. (FIG. 33 is twice as large as FIG. 30.) Deformed Image Memory Background Information Regarding Image Memory Configuration In a raster scan display, the electron beams are ordered along the screen, most commonly. From left to right,
Draw a line from top to bottom. Since display devices use high-speed phosphors to draw motion on the screen,
The generated image disappears shortly after activation, and the image must be reproduced at regular intervals, usually at a frequency of 60 Hz.

In the computer graphic generation system, the reproduction of the image
It is implemented by utilizing a special playback buffer, usually an image memory implemented using solid-state RAM technology. Each storage location in the image memory has a one-to-one correspondence with a respective location on the screen. In that case, the image memory is visualized as a two-dimensional array of storage locations addressed by the x and y addresses. These addresses are generated in synchronization with the electron beam scanning operation. Assuming that the number of horizontal pixels on the screen is M, the x address typically increases from 0 to M-1. When M-1 is reached, the x address value is re-initialized to 0 and the y address is incremented. Y starts at 0 at the top of the screen and ends at N-1 where N is the number of rows from the top to the bottom of the screen. M and N are often in the range of 512 for raster CRT displays, but are in the range of 1280 for color mosaic displays.

Each memory location contains data used to operate the CRT's electron gun to some activation level corresponding to the desired light intensity on the surface of the screen. The data is typically 8 bits for 256 gray levels for each red, green or blue gun. Thus, each storage location can have a depth of 24 bits. In color avionics systems, the number of bits is fairly small, typically three, to indicate any one of the eight colors that may appear on the screen at any one time.
It is a bit. These three bits pass through a look-up table, which decomposes the specified color into its constituent primary color components. The ratio of red, green and blue only needs to be selected with an accuracy of 8 bits.
A very wide palette of colors is obtained on which to base one color choice.

Traditionally, image memories have been implemented using static or dynamic RAM devices. The bandwidth of the RAM is increased up to the video speed by the tracking shift register. Words 16 pixels wide
For example, read into a 16-bin wide shift register, then RA
Clock output at full speed of shift register, much faster than M access time. Today, discrete methods are implemented on a single integrated circuit called video RAM, or VRAM. This chip is very efficient in implementing an image memory design and is used in a preferred implementation. Individually, these chips have sufficient storage capacity to handle the full screen for many applications.

The shift register allows the memory to be accessed by the computer while the data is being read into the CRT, or display. This allows the read cycle and the write cycle to be mixed. While the image is shown, the image can be extracted. However, in real-time applications where a large amount of movement is depicted,
Updating the image during screen playback may also result in visual artifacts. Mixing old and new data on the screen can cause gaps in the image and unusual beat frequencies in the image. To avoid these problems, a ping-pong image memory configuration (FIG. 34) is useful. In this configuration, one memory buffer is used to regenerate the screen while all other image memory buffers are fully allocated to the graphics generator for extraction. The generator notifies the hardware synchronization circuit when the writing of the image is completed. This circuit swaps the buffer during vertical blanking when video is erased from the display. As a result, the complete frame of the image can be displayed at any time without any visible interruption of the content of the image.

Preferred Embodiments of Image Memory In the present invention, any conventional image memory configuration can be used. It does not matter whether you use a dynamic memory chip or a RAM chip. The differences are in the depth of the memory and the allocation. Partial pixel data is stored in addition to normal colors, priorities, and other attributes. The modified image memory is conventionally accessible to the graphics generator and is used to reproduce the display as well. A low cost read / write mixed image update scheme may be employed, but a ping-pong configuration may be selected instead. In a preferred implementation using VRAM (FIG. 34), the ping-pong image memory is 512 × 512 pixels. Each pixel has a depth of 12 bits, of which 6 bits store color and the remaining 6 bits store partial pixel data. That is,
Three bits are used for x subaddressing and three bits for y subaddressing. By storing such data in this manner, the data can be read from the modified image memory for processing by the beam forming apparatus. That is, it is a method or process directed at the read cycle. FIG. 35 shows a deformed pin portion or a pong portion of the image memory.

The attached FIGS. 34 and 35 illustrate this embodiment. For more information on image memory technology, see JDFoley,
A text written by A. Van Dam, "Fundamentals of Interacti
ve Computer Graphics, July 1984, pages 129-135.

Preferred Implementation of the Beamformer Output Section The block diagram in FIG. 36 shows the data path for processing a 3.times.3 kernel to generate a color dot output for an RGGB quad flat panel. The center pixel of the 3 × 3 matrix is expanded to four outputs to generate one output for each color dot in the quad array. This allows the 512 × 512 full field memory (FFM described in the previous section) to be extended to drive a 1024 × 1024 quad green flat panel display. This block diagram can be easily extended to obtain other processing kernel sizes. Alternatively, it is possible to rearrange the block diagram so as to drive another pixel pattern.
The condition required for a 3x3 kernel is that the pixel data from the FFM must be scanned such that the output of the FFM is always one row and one pixel ahead of the pixel address currently being processed. Is represented. The 3x3 processing kernel requires that nine description items be presented for each pixel. This delays the pixels by one row and two rows, respectively.
Implemented by two 512 x 12 bit FIFOs. For each row,
There are three 12-bit registers that are loaded sequentially. Thus, a new pixel from each row loads the first register while the second register receives the previous contents of the first register and the third register stores the contents of the second register. Is loaded. Thus, the nine registers generate, at the output terminal, the contents of a 3 × 3 matrix for processing of a given pixel. The upper three registers hold three pixels from the next data row, the middle three registers contain three pixels from the current data row, and the lower three registers contain three pixels from the previous data row. Includes one pixel. It must be remembered that the image memory always precedes the processing kernel by one data line.

The 12-bit data per pixel can include any combination of color, intensity, and sub-pixel addressing. A preferred combination is a 6-bit color code, 3-bit x-sub-addressing, and 3-bit y-sub-addressing. As a result, the beamformer produces 64 color and intensity combinations, equivalently 512 × 512.
4096 × 4096 addressing can be obtained from the FFM.
Another desirable bit allocation is 4-bit color, 4-bit intensity, and x, y sub-pixel addressing of 2 bits each. This implementation of the beamformer can support two sets of bit assignments with mode bits.

The leftmost 8K × 16 bit memory set contains pixel data
Perform initial conversion to RGGB component output. Memory contains 12 bits from pixel data and one mode bit
Requires 13-bit addressing. Memory is RAM
Base or PROM base may be used. The contents of the memory are loaded with the color dot intensity corresponding to the location, along with a processing window for that pixel's data at each of the nine matrix locations. That is, each memory will usually have different contents. For example, the topmost memory contains the transformations for pixels that are one row ahead for y and one pixel ahead for x. As mentioned earlier, pixel location and sub-pixel addressing compensate for the distance of a pixel from the center of each color dot in calculating how much a pixel contributes in color to the final output of each color dot. I do. The 16-bit output of this first memory bank contributes 4 bits of intensity for each color dot. In the case of a black color code, all color dots output an intensity of 0,
For a white color code, each color dot contribution varies according to the distance to the center of the color dot being generated. The distance from the two green dots is
These two green outputs are often different because they usually differ with respect to the center of the 3x3 processing window. The distance discussed here is the distance from the center of the individual color dot being generated to the location of the subpixel in the pixel being processed (a total of nine pixels are processed in parallel to generate four color dots). ing).

The 16-bit output of the first set of memories is four 4-bit (R,
G1, G2 and B). The second set of memories, shown in the block diagram, combines the intensity data row by row such that each memory outputs an intensity contribution for each of the three rows, for each of the four color dots. You. The second set of memories in this implementation is 8K × 8 bits each, in which case the 13 address bits consist of three 4-bit strengths and one mode bit. The mode bits allow the two anti-aliasing strength combination algorithms to reside. The PROM can select the maximum intensity and either linearly add the contributions or perform a non-linear addition on the intensity components. Any of the combination functions can be loaded into these memories. The three rows each contain four of those memories. Each memory receives three 4-bit strengths from the first set of memories for a particular color dot. That is, each row will output the strength contribution for each of the four color dots. Note that only four bits of each 8-bit memory are required (8K.times.4 bit memories are not readily available). The third set of memories combines the output for each of the three rows and provides a final output for each color dot. Generate output.
Each memory receives three 4-bit strengths from each row for a particular color dot. That is, all three red color dots (Ra, Rb, Re) are sent to one memory (along with the mode bits), where their contributions are combined to produce the final red color dot output. The combination algorithm is programmable, but is usually considered the same as used in the second set of memories. The two sets of green and blue intensities are also sent to separate memories to generate the final G1, G2 and B outputs. In this application, only 4 bits of intensity are generated, but in practice, 8 bits are available from the 8K × 8 memory.

For flat panel displays using different color dot arrangements, the architecture is easily rearranged. In case of RGB diagonal pattern or RGB delta pattern,
The color dots can be similarly arranged in groups of four, but the color dots move around. However,
At most three types of patterns appear (RGBR, BRGB and GBR
Since G), the architecture is modified to have two mode bits instead of one. Thus, each memory must have 16K storage locations, not 8K. In that case, the mode bits change each clock cycle to select the storage area reserved for each color dot pattern. However, the final output will be based not only on the color but also on the position (in the case of RBBG quad panels, the color assignment does not change based on the position), ie the upper output (out of the four) is the RGBR pattern. , But drives B or G for the other two patterns. Therefore, also in this case, regardless of the arrangement of the color dots,
The 512x512 FFM can be extended to fill all the color dots in a 1024x1024 flat panel.

[Brief description of the drawings]

FIG. 1 is a diagram showing a difference in the shape of two lines in which a line with aliasing is jagged and a line in which aliasing is prevented is smooth and continuous. FIG. Shows the relationship between the input of the beamformer element, which is the impulse data from the system, and the output, which is the point spread function of the beamformer convolved with the impulse input, and as a result of centering the partial pixels,
FIG. 3a illustrates the increased accuracy of the positioning of color energy with respect to the display device. FIG. 3a utilizes a write cycle aliasing prevention method performed by the graphics processor while the image is being rendered and written to the image memory. FIG. 3B is a diagram showing a conventional image processing architecture. FIG. 3B is a diagram showing a display device driving system of the present invention in which a beam forming device is added to a deformation system which is a read cycle anti-aliasing system in the present invention. FIG. 4 shows the position of a partial pixel as an example of where in the pixel the center of gravity of the luminance can be commanded according to the invention. FIG. 5 illustrates the concept of the location of a sub-pixel within one pixel in an image memory, FIG. 5 provides not only the standard x and y address values to be used when addressing the image memory, FIG. 5a shows a modified graphic processor used in the present invention which also supplies partial pixel address data, i.e., fractional part address data. FIG. 5a not only stores 71 such as color data or attribute data, but also generates modified vector data. FIG. 6 is a view showing a modified image memory which also stores partial pixel address information 72 obtained from the imager. FIG. 6 shows an example of a slide window used for processing of the beam forming apparatus. It shows the operation method of the beam forming apparatus to take out and execute appropriate processing to generate a single output pixel, Window goes line by line from left to right,
FIG. 7 shows that the data is moved from top to bottom. FIG. 7 first executes color processing to classify data according to the next output pixel (only in the case of a color mosaic display device). Perform a brightness lookup based on the Gaussian profile for each active pixel in the window,
Third, a block diagram showing the internal details of the beamformer, comparing all shades to determine the maximum shade, and finally outputting this shade to the display, FIG. And FIG. 9 show an example of color processing showing a slide window in the image memory including a red line and a red line. FIG. 9 shows an example of color processing of color classification executed for the next green output image. FIG. 10 is a diagram showing an example of color processing of color classification performed on the next red output pixel. FIG. 11 is a Gaussian profile in a sliding window used for a shaded lookup value. FIG. 12 to FIG. 17 show one point passing through the slide window in the image memory, and FIG. 18 shows an example of a slide window representing six consecutive time frames. Figure FIG. 26 shows three points close to each other in the image memory and shows the position of the sliding window in nine consecutive time frames, similar to FIG. 12 to FIG. The figure shows a Gaussian profile based on an 8-step gray shade, showing the luminance profile for a 20 mil wide half-amplitude line drawn using 8-step gray shades, assuming a pixel spacing of 6.7 mils. Figure 28 is a diagram of a Gaussian profile based on a six-step gray shade, showing the luminance profile for a 20 mil wide half-amplitude line drawn using six steps of gray shades, assuming a pixel spacing of 6.7 mils, FIG. 29 shows a luminance profile for a 20 mil wide half-amplitude line drawn using four levels of gray shades, assuming a pixel spacing of 6.7 mils. Diagram of a Gaussian profile based on a shade, FIG. 30 shows a vector drawn in an image memory, FIG. 31 is a block diagram of a vector generator, and FIG. 32 is used with a beamformer. FIG. 33 is a block diagram of a vector generator, FIG. 33 is a diagram showing a vector drawn using the modified vector generator of FIG. 32, FIG. 34 is a diagram showing an embodiment of a ping-pong image memory configuration, FIG. 35 is a diagram showing a modified embodiment of a pin portion or a pong portion of an image memory. FIG. 36 is a block diagram showing an embodiment of a data path for processing a 3 × 3 kernel for generating a color dot output. FIG. 20a ... deformed figure generator, 21a ... deformed image memory,
22 ... display device, 23 ... beam forming device.

Continued on the front page (72) Inventor Brent Hans Larson United States 85306 Granddale West, Caribbean Lane, Arizona 6134 (56) References JP-A-63-198094 (JP, A) US Patent 4843380 (US, A) (58) Field surveyed (Int. Cl. ⁷ , DB name) G09G 5/36 G06F 3/153 G09G 1/16 G09G 5/02

Claims (4)

(57) [Claims] An improved graphics system which applies an anti-aliasing filter to the output of a partial pixel image memory arrangement to improve the image quality of a raster display device, wherein a standard integer x to be used when addressing the image memory. A modified figure generator having output means for supplying not only address data and standard integer y address data but also partial pixel address data, ie, fractional part address data; and integer address data supplied by said figure generator. An input terminal connected to receive integer address data and partial pixel address data from an output means of the graphic generator, for storing partial pixel address data, i.e., fractional part data; Modified image memo with output means for supplying Having input means connected to the output side of the image memory and output means connected to the input terminal of the raster display device, receiving impulse data from the image memory and transmitting the impulse stator to a display device. Before expanding to a point spread function (ie, Gaussian profile)
A beam forming device for adjusting the center of gravity of the point spread function using the received partial pixel address data; and a display device having an input terminal and receiving an image to be displayed from the beam forming device. An apparatus comprising a combination of 2. A method for improving image quality of a raster display device, wherein not only standard integer x address data and standard integer y address data to be used when addressing an image memory, but also partial pixel address data, that is, a fractional part address Providing a modified graphic generator having output means for also supplying data; and a storage device for storing integer address data and partial pixel address data, ie, fractional part address data, supplied by said graphic generator; Providing a modified image memory having an input terminal; and connecting an input terminal of the image memory to receive integer address data and partial pixel address data from the graphic generator; Has output means for supplying partial pixel output Having input means connected to output means of said image memory, receiving impulse data from said image memory, and diffusing said impulse data to a point spread function (such as a Gaussian profile) before transmitting said impulse data to a display device; Providing a beam forming device that adjusts the center of gravity of the point spread function using the received partial pixel address data; and having an input terminal to receive an image to be displayed from the beam forming device. Providing a display device. 3. An improved graphics system which applies an anti-aliasing filter to the output of a partial pixel image memory configuration to improve the image quality of a color mosaic display device, wherein both integer address data and partial pixel address data are stored. A graphic generator having an information output terminal for supplying and including a vector generator and an image memory; a color mosaic display device having an input terminal and receiving an image to be displayed from the image memory; Connected between an output terminal of a color mosaic display device and an input terminal of the color mosaic display device, receives the impulse data from the graphic generator, and modifies the image received by the color mosaic display device to improve image quality. Diffusion luminance function, which is the point spread function of the beamformer convolved with the incoming pulse data A beam forming device for providing a file output, wherein the image memory stores a plurality of pixels for the image, including both active pixels and inactive pixels, wherein each active pixel of the image memory is stored. Is a device that is extended to a luminance profile by the beam forming device before transmission to the color mosaic display device. 4. A method for improving image quality of a color matrix display device, comprising: a graphic generator including a vector generator and an image memory for providing an information output including both integer address data and partial pixel address data. Providing; storing a multi-pixel matrix image in the image memory; providing a color matrix display device; receiving information output address data from the graphic generator;
A beam forming device is provided between the image memory and the color matrix display device, which expands each active pixel of the information data from the image memory to a brightness profile which is a point spread coefficient of the beam forming device before displaying. And a process.