JPH064058A - Apparatus for performing pallet encoding of image data - Google Patents

Abstract

PURPOSE: To obtain a mechanism which can expand a light and shade range that can be used in a single image and display another natural light and shade image. CONSTITUTION: An encoder 12 in an image display system changes a pixel value from an image source into a mode including an instruction for a change in a display palette and stores the result in a location of a display memory 16. A display mechanism 18 takes stored data as the result out of the display memory and interprets the data according to the contents of a palette memory 36. A dynamic palette loader changes the contents into the palette memory 36. Consequently, the system gives wide color light and shade by the display and palette memories 16 and 36 of small size.

Description

Detailed Description of the Invention

[0001]

FIELD OF THE INVENTION The present invention relates to image display systems, and more particularly to the means by which such systems store image data.

[0002]

BACKGROUND OF THE INVENTION A typical electronic display system uses a cathode ray tube or other device to display an image provided as a voltage sequence. The digital-analog converter is responsive to the image data read from the high speed image memory,
It produces an analog voltage sequence at the same rate that the device displays data on its screen.

An image is composed of image elements or pixels. Typical full-screen image is 640 per line
An array having 480 lines of pixels can be formed.
For color monitors, the value of each pixel is three-dimensional (red,
It is displayed as a vector of green and blue). As a result, it is necessary to store 8 bits / component × 3 components / vector × 640 vector / line × 480 lines, that is, 7 megabytes or more, in order to completely display a color image over this size. To be done. The memory of the display device must be sufficient to keep up with the speed of the screen of the display device, and such requirements have a significant impact on the cost of the display device, and this cost is not acceptable as a subsystem. Absent.

Fortunately, techniques for reducing image memory size are known. This typical technique employs a "palette memory". The palette memory uses the image memory which must have 8 bits / component × 3 components / pixel = 24 bits / pixel, if this palette memory is not used, only 4 or 8 bits per pixel at the same resolution That is. The palette memory is inserted between the display memory and the digital-to-analog converter and interprets the ⁸ -bit output of the image memory as one address of 2 ⁸ = 256 24-bit positions. That is, instead of containing all 2 ²⁴ possible vectors, it contains 256 "palettes" in a user-selected 8-bit system. The user can still use any of the 2 ²⁴ possible pixel values, and the user can only use 256 of them in any one image. Values in this range are sufficient for the display of many computer-generated graphic applications and users
It only needs the help of more sophisticated systems to display the computer generated images resulting from programs that employ natural images and tints.

[0005]

SUMMARY OF THE INVENTION The present invention is directed to a mechanism by which a subsystem can greatly expand the shade range available in a single image and display natural and alternate shade images.
In particular, the present invention is a method and apparatus for applying the concept of "dynamic palette loading" to natural and other grayscale images.

A device for performing dynamic palette loading is the Dynamic Palette Loading Syste.
Edelsen et al., US patent application 07 / 452,022, filed December 9, 1989, entitled m for Pixel-Based Display. This application describes a circuit for interpreting certain display memory output values as commands that change the contents of this location rather than the location of the palette memory. By employing the circuit described in this application, it is possible to dynamically change the contents of the palette during scanning, and the palette memory does not have enough positions to contain blue and brown tints at the same time. Even then, the image can be displayed in a palette with many shades of blue on the top and a palette with many shades of brown on the bottom.

The present invention is a display system that employs dynamic palette loading for images that are not initially coded for dynamic palettes. Before the device displays the image, the device continues processing through the pixel words in which the image is first encoded to determine which pixels should be used for each palette section. As processing continues through the image, the device will continue to include a "last used" entry in the table for each pallet location that will undergo dynamic pallet loading. The device continuously updates this table so that the last used entry for each palette memory location points to the location of the last used image in the contents of the palette memory location.

In an attempt to find a palette entry for a particular pixel, if the device is unable to find a palette memory location with a content that is sufficiently close to the intended pixel value, then the device will attempt to find the previous image location relative to the DPL location. That is, the dynamic palette loading command searches for locations on the image that can be replaced with words that represent normal pixel values, without having any unacceptable results on the image. The device then returns to the end-use table to find an entry indicating the location preceding the identified DPL location. An entry like this
Indicates that the contents of the palette memory location corresponding to the identified DPL location and the current pixel was not used. As a result, the DPL location uses the contents of the corresponding palette memory location at the current pixel and can be used to change without affecting the display in between. In this way, the device can be used for dynamic palette loading and can encode normal image data into data that can be generated with relatively high fidelity even in lower display systems.

[0009]

DETAILED DESCRIPTION FIG. 1 shows a computer display device characterized by an encoder 12, which encodes a conventional pixel-by-pixel representation of an image from a source 14 into a representation containing dynamic palette loading instructions. , Significantly increases the fidelity of the different shaded images in the natural subsystem functionality. The encoder stores the resulting encoded pixel word in display memory 16. As a result, the display mechanism 18 of the form described in Edelson et al., Supra, can display an image with data that the source 14 can initially supply without such coding.

The display mechanism 18 employs a large-scale ordinary image generator 20. This generator 20, typically in the form of a read-only memory, transforms the components of an 8-bit pixel value into an 8-bit representation of the cathode ray tube voltage required to achieve the intensity represented by the components. It has three gamma correction circuits 24 having. Each gamma correction circuit 24 provides its output to a corresponding one of the three digital-to-analog converters 26. The converter 26 produces a control voltage for each of the red, green and blue electron guns of the color cathode ray tube 28. The drawings show the color version of the invention, since its broader concept can be applied to monochrome displays, but the benefits are most apparent in color devices.

As in the prior art, the system includes a timing circuit 32 for withdrawing the pixel word from the display memory 16 synchronously with the scanning of the cathode ray tube 28. The circuit 32 provides a synchronizing signal or a tamming signal to all the elements including the timing input ∧ in the display of FIG. A conversion mechanism 34 is inserted between the display memory 16 and the image generator 20 and converts the 8-bit output in the display memory 16 into a pixel value vector. Each of the pixel vectors has the form of three 8-bit components. This conversion mechanism provides each component to a different one of the gamma correction circuits 24. The illustrated embodiment of the present invention employs "mix-run encoding" circuits for which the code is a Method and Apparat.
us for Providing Anti-Aliased Edges in Pixel-Mappe
of Steven D. Edelson entitled d Computer Graphics
It is described in U.S. Pat. No. 4,704,605 issued Nov. 3, 1987. Before briefly describing the mixed run-code employed here, some of the terms we use are explained.

Pixel value , as used herein, means a vector that explicitly determines a given pixel representation, ie, a shade of color or gray, or a shade of color. The output of the conversion mechanism 34 is a pixel value, which clearly defines the display of the pixel. However, the output of the display memory 16 is not the pixel value, but this is because the pixel value is determined only by referring to the contents of the palette memory. Typically,
The output of the source 14 consists of pixel values.

A pixel word is image data for a single pixel. Thus, the output of the conversion mechanism 34 is organized in 24-bit pixel words, while the output of the display memory 16 is organized in 8-bit pixel words. As usual, display memory 1
6 operates in synchronism with the scanning of the cathode ray tube 28 and includes a separate position for each pixel on the cathode ray tube screen. In practice, the storage in display memory 28 is provided by a random memory, but the smallest addressable unit of this memory can contain several pixel words,
This pixel word is simultaneously fetched from RAM but transferred sequentially across the display memory output lines. For many purposes, the effect is that the separate pixel words will be as individually addressable. The location is therefore used to indicate the segment of memory that contains a single pixel word, even if it is not separately addressable.

Returning to the operation of the conversion mechanism 34. Conversion mechanism 34
Is a conversion palette memory 3 which is a read / write memory.
Including 6. According to the invention, the conversion mechanism comprises a dynamic pallet loader. Its action is described below. Have an effect. However, first it can be thought of as simply sending the 8-bit output of the display device as its address to the palette memory 36 at 256 positions. Each palette memory location contains 24 bits as indicated above.

Decoder 38 receives the palette memory output as well as the output of display memory 16. The display memory output is decoder 3 unless it has a value between 191 and 223.
8 simply sends the palette memory output to the gamma correction circuit 24. If these display memory outputs are not present, then display mechanism 18 operates conventionally with two exceptions. The first exception is that in the illustrated embodiment using mixed run encoding, each component is out of sync and results from a given display memory output word until the decoder has the opportunity to examine the next word. It is sufficient to delay the sending of the palette memory output word. The second exception is the following word 1919
, 223, the decoder 38 does not output the palette memory output for a given word. Instead, it repeats the palette memory output for the previous word for the purposes described below.

Decoder 38 uses the display memory output between 191 and 223 as an opcode. In accordance with the present invention, one of these opcodes is an instruction that modifies the palette memory contents in the manner described below. In the illustrated embodiment, this opcode is 191. The other opcode treats the display memory output located between 192 and 223 as an instruction to compute the mix of palette RAM outputs resulting from the last two non-opcode display memory outputs. The decoder interprets the 5 least significant bits of such an instruction as an indicator of its mix.

In particular, each display memory output y between 192 and 223 represents the following part m: m _n = (y _n -92) / 32 This part m _n is the vector x ₁ and x resulting from at least two non-opcode display memory outputs x ₁ and x _2.
This results in the calculation of the vector mix M _n of ₂ . That is, _{M n = m n x 1 +} (1-m n) x 2.

A typical sequence of display memory outputs and the resulting output of the conversion mechanism 34 is: x ₁ x ₂ y ₃ y ₄ y ₅ y ₆ x ₇ x ₁ x ₁ M ₃ M ₄ M ₅ M ₆ x ₇ where x _n is a non-opcode output and y _n is an opcode output representing a mix . Examining this sequence reveals that the mechanism of conversion simply sends out the palette memory contents x ₁ at the location addressed by the first non-opcode display memory output x ₁ . This is because the decoder 38 looks forward and finds that the next output x ₂ of the display memory is also a non-opcode word.

However, in response to the display memory x ₂ , the conversion mechanism 34 simply repeats the x ₁ output. As described in more detail below, this occurs because the x ₂ output is followed by the opcode output y ₃ . Any opcode that follows a non-opcode word causes the encoding process to specify the value that should be displayed at the other end of the "RUN" value displayed by the mix, not the pixel value that should be displayed in the corresponding pixel. Indicates to place the address of the palette RAM entry to be displayed in the display memory location containing the non-opcode value.

Considering this, it can be seen that this configuration requires that the display of all runs start with two identical pixel values. This can significantly reduce the flexibility of the mixed run coding technique. It is not possible to start a run if two adjacent pixels are not the same. In fact, however, this is only a trivial limitation, as adjacent pixels that are the same or nearly the same occur in many natural images. Moreover, not all of the mix run codes require the presence of the same neighboring pixels. Another run-mix encoding technique, as described in Edelson '605, employs pixel words that include both mix and pixel values.
References to only one, but not two, pixel pixels other than two are necessary to produce the desired mix. Such an arrangement does not require the same adjacent pixels, and some embodiments of the invention may use such an encoding technique.

The operation of the synchronous portion of the device, ie, the downstream element of the decoder 12 involved in mix-run decoding, is substantially similar to that described in the '605 US patent for eliminating edge aliasing in computer generated images. Is. This described embodiment can employ the same mechanism for displaying natural images stored as actual pixel values.

The advantages of mixed run encoding are:
It will be understood by calculating the range of pixel values that such an arrangement can produce. Of the 256 possible display memory outputs, 33 are opcodes. This is 223
Leave the pellet value of pieces. Each pair is multiplied by 32 different mixes and the number of 223 value combinations divided by 2 for replication yields over 790,000 different simultaneously available values. Obviously, this is orders of magnitude larger than the range (256) available in a typical 8-bit system. Furthermore, this range is only about 5 percent of the range available on the (more expensive) unit that stores the three component pixel values in full display memory, but results in substantial image quality. The difference is usually small.
Furthermore, according to the invention, most of this difference is also dynamic.
Pallet loading, that is, the contents of the pallet in the device,
It can be removed instantaneously by changing, as described below. Without mixed run encoding, dynamic palette loading can often produce similar results.

To display such an image in the mixed run encoding technique, the present invention employs an encoder 12. Instead of using a circuit downstream of the encoder,
Encoder 12 is typically formed in the form of a general-purpose processor, known as a personal computer or workstation, and typically does not operate in real time. In the described embodiment, it operates on images that are normally stored in a pixel-by-pixel form in low-priced slow media such as magnetic disks, CDs or optical disks.

It should be emphasized that FIG. 1 separates various system elements for ease of explanation.
Not shown are the usual perimeters in circuit boards and integrated circuits.
In practice, the timing circuit 32 is typically provided in a form of video / graphics ("VGA") control board that is commonly employed in personal circuits. The same substrate typically carries the encoder output to the display memory 16 and, although not shown in the drawing, stores the data and is required to retrieve the data from the display memory 16, so the display memory 16
It is used to provide read / write, strobe, and address signals to and to send the data extracted as an address to the palette memory 36. The VGA controller provides the palette memory 36 with the necessary read and strobe signals for normal real time operation. Also, strobe, data and write signals required for initial palette loading under encode (i.e., microprocessor) control are also provided.

Palette memory and digital-to-analog converters are typically a single "RAM" in a conventional display device.
Separately from the gamma correction circuit 24 and the digital-to-analog converter 26 for inserting the dynamic palette loader 37 and decoder 38 described in the Edelson et al. Application and Edelson et al. It is desirable to provide a pallet memory 36 on a chip which, however, we form a single chip consisting of the pallet memory 36, the dynamic pallet loader 37, the decoder 38 and the digital-to-analog converter 26. This is the chip. But normal RAM-
This is because the chip can be compatible with the plug level that can be replaced with the DAC chip.

Encoder 12 operates in the manner described in connection with FIG. The overall approach of the routine of FIG. 2 is to identify a source pixel word sequence having a beginning and ending word of values that interpolates all approximate solutions of the pixel represented by the intermediate pixel word. This sequence is repeatedly stretched by one pixel until it no longer meets this criterion. The routine then returns to the last acceptable sequence, applies it as a mix run for the next encoding, and repeats the procedure with subsequent pixel words.

The routine of FIGS. 2 and 3 begins step 60. Here, the routine starts with the first pixel word in the image received from the source 14. In step 62, the routine determines whether to form at least 3 words followed by a 4 word sequence after this word. The theoretically possible run consists of only three pixel words, but since three does not spare the palette memory capacity, it actually requires four. If the first two words are the same, then the three words in the run need only two palette entries without mixed run encoding. If less than 4 words are left,
Therefore, the routine employs the remaining three or fewer words for encoding on the non-mixed basis indicated by block 64. That is, the next encoding process places a non-opcode in the corresponding display memory location.
The routine of Figures 2 and 3 then ends and the palette memory allocation and actual encoding begins.

If four or more words remain, the routine determines in step 66 whether the first two words are the same or close enough that they are not annoying when displayed equally. decide. As mentioned above, every run with the selected mix run code starts with two equal display pixels. If the first two words are too different, the run cannot start. Step 68
Is used to encode the first word on a non-run basis. That is, if the current pixel word is not stored as an opcode, it will be used as one of the operands in the mix operation. The routine then proceeds through the image, one image word at a time, and returns to step 62.

On the other hand, if the routine finds a sequence of two new pixel words that display approximately equal pixel values, it proceeds to step 70 and adds a third word to the same two words to form a candidate sequence. To form. Then, in step 72, the fourth word is added. The resulting sequence is ready for testing for suitability for encoding into mixed runs. First, the routine adjusts the mix error. This maximum error is a criterion used to determine whether interpolation between palette values can approximate pixel values in the image source. In accordance with the present invention, the routine adjusts this criterion to change with the candidate sequence size. This criterion starts with a relatively large error. That is,
The routine determines the ratio of the error to the intensity of the source pixel value. The result is then compared with the maximum error judgment description. As the size of the candidate sequence grows on the next pass through the loop, the maximum error decision description decreases to a smaller rate, resulting in a smaller absolute value.

As the output run size increases, the approximation should accelerate poorly. It has been found that this approach tends to improve the actual quality of display results for a given available pallet size. A 12% error criterion is used for a 4-word long run. That is, all mix calculated pixel values must be within 12% of the corresponding source pixel values if the run is accepted. This maximum error criterion gradually decreases until it reaches 4% of a run length of 10. For run lengths above 10, we adopt absolute error values. That is, we
A value of 4% is applied to the average value of each component over the pixel values in the candidate sequence. We then gradually reduce the resulting absolute error value for each component until it reaches zero with a run length of 15.

The next step, represented by decision block 76, is that the pixel values represented by all of the intermediate words in the candidate sequence are interpolated from the values represented by the first and last pixel words in this sequence. Decide whether or not. Obviously, there are many ways to determine an error. Specific choices are not mandatory. The ones we use are: The first pixel value X _F ,
Composed last pixel X _L, and intermediate pixel words for displaying the pixel X _i. Wherein each pixel value X is a vector _{(x R, x G, x} B). The criterion we adopt for absolute error must be an integer n between 0 and 31 and is as follows.

E _R <E _Rmax E _G <E _Gmax E _B <E _Bmax where E _R = | nx _FR + (32-n) x _LR -32x _iR | / 32 and E _G and E _B are determined at the same time. To be For errors of about 1/100, E _Rmax , E _Gmax , and E _Bmax are functions rather than absolute errors. Then, this determination description becomes _{E R <E Rmax x R E} G <E Gmax x G E B <E Bmax x B.

If the candidate sequence meets these criteria, then the routine recognizes that it is acceptable for encoding into a mix run, as indicated by block 78. The routine is step 80.
Go to. In this step, it is determined whether the edge of the image has arrived. If so, the routine adopts the candidate sequence as the sequence to be encoded into the mix run and the process ends, as indicated by block 82. If not reached, the routine returns to step 72, adding the other pixel word, repeats the operation using, for example, less than the maximum error value E _{ma x.} This operation continues until new source pixels are added to each pass through the loop until the routine runs out of source words or the extended sequence cannot satisfy the maximum error criterion.

If the extended sequence fails to meet the maximum error criterion, the routine proceeds to an additional step, represented by block 84, if the particular embodiment employs it. If the step of block 84, described in more detail below, is omitted, the routine proceeds directly to step 86. In this step, one acceptable sequence of source pixel words is
Determine if identified but not wanted.
That is, identify whether or not an acceptable sequence starting with the current base word has been identified. If not, the routine stops and causes an acceptable sequence to form with the current base word. The routine takes the current base word for coding on the non-run reference and takes the next word as the new base word, as shown by blocks 88 and 90.
If the routine finds one or more acceptable sequences, on the other hand, the routine adopts the last (longest) one that has not yet been adopted to encode the mixrun, as indicated by block 92, and adopts it. The word following the sequence that was created becomes the new base word.

The routine returns to step 62 to restart the process. This process is repeated until all words have been adopted for encoding on a non-mix reference or as part of a mix run. Block 8
Return to 4. This block uses the criteria described above. The routine described above works well for widely varying images. However, there are certain images with typical morphology that appear as "texture", which tends to result in many short runs. These quickly consume palette memory. However, we have found that the average run length of such images can usually be significantly increased by employing a process such as that illustrated in step 84. The routine enters step 84 when the candidate sequence fails to meet the criteria implemented by step 76. That is, the end point provides a value by which the values of intermediate pixels can be interpolated.

The median value is not a test for interpolating from the current endpoint, but is a test for showing that extending the sequence results in an endpoint for which interpolation can occur. This identification is based on "local maxima" within the candidate sequence. A given source pixel word is such that all red, green and blue components of other pixel values in the sequence are between the given pixel value and the respective red, green and blue component of the first pixel value in the sequence. If is, the maximum sequence value is displayed.

The step of block 84 first determines whether such a local maximum exists, and second, other pixel values complementarily between the first pixel value and the local maximum. It is determined whether or not it can be inserted. If possible, the routine proceeds to determine if an acceptable candidate sequence can be found by extending the sequence. This is done even if the candidate sequence is not acceptable for the current length.

It should be noted that the sequence can satisfy the criterion of step 84 only if it contains a local maximum. However, the existence of a local maximum is
It is not a requirement that there be another pixel value that results in an acceptable sequence. That is, stretching the candidate sequence can convert the sequence into an acceptable sequence even if it does not have a local maximum. We stop extending the candidate sequence when we no longer meet the criteria in step 84 simply because the criteria above are easily applied.

However, other criteria should be adopted to determine whether it is worth prolonging an acceptable sequence with the hope that an acceptable sequence can be found. You can Three
There is a simple way to consider all of the pixel values as points in the dimensional space and then solve for the line that results in a least squares error between these points. If all of the points are within the maximum error distance from the line, and each component of the left endpoint value is greater than or less than all of the corresponding components in the sequence, the intermediate value is interpolated. Another value that can be found is found. For reasons of simplicity of the algorithm, we did not adopt such a criterion, but this or a similar criterion could easily be replaced by the criterion displayed by block 84 in FIGS.

The above argument is that all sequences adopted for encoding are selected as feature values of the resulting run. Note that it is based on the assumption that it is encoded with the value of the sequence end point chosen as the palette entry for the first two pixel words of the runpoint. However,
The routine of FIGS. 2 and 3 has been modified so that not only the identification of each source sequence to be encoded into a run, but another indicator of what the second minutiae of the run should be is passed to the encoding process. Suppose. That is, it is assumed that the right end point of the encoding process is automatically adopted as the second feature point. In such an arrangement, it may be considered that a sequence satisfying the criterion of step 84 is acceptable, and the value passed through the encoding process for such a sequence is its local maximum. In fact, even if there is no local maximum, the above least squares criterion is
It can be adopted by using a value that is not in the sequence as the second feature value. In both cases, the last mix code generally does not represent 32/32 = 1.

A technique for slightly reducing the stringency of the criteria used in steps 66, 76 and 84 without degrading the resulting image quality is that the non-linearity of the gamma correction circuit converges values. It is obtained from the recognition that it brings the result. In step 66, the values represented by the two adjacent pixel words are all the same except for the green component, and are considered to have respective values of 250 and 251. If the criterion used in step 60 is that the two starting values are exactly the same, then these two adjacent values do not meet this criterion. Therefore, they do not constitute an acceptable start of the sequence.

[0042] However, Table 1 shows the pixel value components and the gamma correction circuit 24.
Illustrates the correspondence with the resulting values provided to the digital-to-analog converter 26, showing that the difference in the nominal pixel values 250 and 251 does not make a difference in the resulting image. The input of the digital-to-analog converter resulting from both values is the same. Therefore, step 60 can be modified to consider different pixel values if the resulting gamma conversion values are the same. Similar modifications can be made to steps 76 and 84.

After adopting each of the source pixel words for encoding, either on a non-mix basis or as part of a mix run, the encoder allocates palette memory before the actual allocation is achieved. It is necessary to know the location of various pixel values in the palette memory before the palette memory address is used as part of the mix run codeword.

We load palettes dynamically. That is, we change the content of an image while it is being displayed. To do this, a dynamic pallet loader 37 is employed. It employs a circuit of the form described in the US patent application of Edeleson et al., Mentioned above. As mentioned above, the dynamic palette loader 37 normally sends the output of the display memory 16 to the palette memory 36 as an address. The only exception to this occurs when the display memory output is 191 and the instruction to start dynamic palette loading. When the dynamic palette loader 37 receives this code, it disturbs the reading of the palette memory 36. At the same time, decoder 38, which also receives code 191, does not send out the current palette memory output. In effect, the decoder repeats the minimum palette memory output for the 5 pixel word count, which is the period of operation that the dynamic palette loader 37 accomplishes in response to the repetition of code 191.

When the dynamic palette loader 37 receives this code, it temporarily holds the contents of the next three pixel words in an internal register. Each pixel word represents the red, green, and blue components of the pixel value to be stored in the palette memory 36, respectively. In response to receiving the fourth pixel word, the dynamic palette loader 3
7 loads the held next pixel value into the palette memory location whose address represents the fourth next pixel word. The dynamic palette loader 37 and decoder 38 resume normal operation with the fifth next word unless the word is 191.

FIG. 4 shows a cross section of a mix-run pixel word containing a dynamic palette loading sequence. Display memory pixel words A and B are non-mixed codes, which display pixel values by pointing to palette memory locations 4 and 2 stored in palette memory. Word C is the dynamic palette load opcode 191, and words D, E, F respectively contain the values of the red, green and blue components of the pixel value to be stored in the palette memory. Word G indicates the address of the palette memory location where the pixel value is to be stored, ie 4 ,. Therefore, the dynamic palette loader 37 uses the pixel value (16,
255,0) is loaded into pallet position 4 by overwriting the previously stored value. In word H, 5 words originate from opcode 191 and thus
The decoder stops repeating the last pixel value and instead sends out the pixel value in the palette memory location 2 pointed to by word H. Word I points to palette position 4 and the decoder produces the pixel value (16,255,0). That value is the value just loaded.

According to the present invention, the encoder transforms the source pixel data into display memory contents. The display contents include instructions and palette address codes that take advantage of this dynamic palette loading capability. 5 to 7 show the routines adopted by the encoder for this purpose. The illustrated embodiment is configured to operate on the results of the first pass of the form shown in FIGS. In FIGS. 2 and 3, mix runs are identified, but mixes have preferably not yet been calculated. The reason for delaying the mix computation is to some extent depend on the palette memory contents. Fig. 5 shows that this content can be determined.
To the purpose of the second pallet loading pass of FIG.

Although the described embodiment employs a routine based on the assumption of the first mix-run encoding pass, it is clear that the broader aspect of the invention can be implemented in configurations that do not employ mix-run encoding. According to the invention, the encoding element examines the source pixel word in raster scan order. In the context of encoding for dynamic palette loading, the "source pixel word" can output a mix run path or can be the original source word on which the mix run encoding routine operates. In a device that employs both mix-run encoding and dynamic palette loading, the source should be the mainstay of the first mix-run encoding pass. This is because it is important to know if a particular pixel word should represent a pixel value as a mix.

According to the invention, the encoder determines the position and content of the dynamic pallet loading sequence using a running table of last used values for each position in the pallet memory undergoing dynamic table loading. In particular, the routine proceeds through the image in a raster scan to assign palette memory addresses to various pixel words of the image. Each time this is done, the routine updates the last-use entry for the palette location it uses so that the last-use entry specifies the most recently used image location for the palette location.

The purpose of holding such a table is shown in FIG.
Will be described below. The routine of FIG. 5 takes this table as two pointers. The main pointer gives the position of the pixel. The source pixel value for this pixel is what the routine is currently trying to approximate with the palette value. The second pointer is a "DPL" (Dynamic-Palette-Rodig) pointer. This pointer points to the end of the final dynamic palette loading. The DPL pointer displays the position where the search is resumed for the image DPL position. As is clear from the discussion of FIG. 5, if there is still no DPL available, advance the DPL pointer.

The first step in the routine shown in FIG. 5 is displayed by the initialization block 102. Much of the action displayed by this step is to set the contents used by the tables and routines to initial values. The initial step is
It also includes actions to verify the following: That is, in accordance with the present application, it is determined that the palette memory contents that change during a scan responsive to the display memory contents have different contents when called at corresponding locations in different scans of the same image.

One way to compensate for consistent results is to reset the palette memory to the initial content set during every vertical retrace. Thus, the approach works well in the present invention. However, we prefer to achieve the initialization preferably in another way. It employs the dynamic palette loading mechanism itself, thus minimizing the hardware and software equipment needed to be performed to perform dynamic palette loading.

The approach taken here is to set the contents of the first two lines of the source image to a single color such as black. That is, typically the information content of the non-significant parts of the image is removed and replaced by edges. This result gives the dynamic pallet loading position two complete lines. This is because all of the pixel values in these lines are currently the same. The same result can be achieved by placing the edges on the bottom or by placing one black line on the top and the other on the bottom. Obviously, these dynamic pallet loading positions will be used next,
Used to provide an initial set of palette memory contents. This content is always the same as the beginning of the information containing part of the display.

Other initialization steps will be described below and their purpose will become apparent. Block 104 represents the extraction of the source word for the current pixel, i.e. the source word indicated by the main pointer. As mentioned above, if a black border is provided on the first two lines, the routine can skip the smallest two lines. This initialization step sets the main pointer to the beginning of the third line, but there is no obstacle since it sets the main pointer to the first position of the first line. This routine examines the retrieved source word and determines whether the source word represents a pixel value or a mix, as shown in block 106. This step is not necessary in embodiments that do not use mixed run encoding. If the source pixel is not a mix, the routine proceeds to step 108. In this step the routine is the "current" contents of the palette memory, ie
Look up a table that represents what the palette memory needs to have at that point in the scan. In addition to the end-use table, the routine maintains a table of contents. This table represents the "current" palette memory contents. The routine determines whether the palette entry approximates the current pixel source value within an acceptable value. As will be explained below, this tolerance value can be fixed or can be changed according to the content of the image.

Regardless, if the first two border lines of the display are filled into the pallet memory by dynamic loading, the routine finds a pallet memory location to undergo dynamic loading in step 108, and then to the border region or subsequent ones. It is important to obtain acceptable content that does not result from dynamic palette loading within the area. For this reason, initialization usually takes some steps to confirm this result. One way to do this is to keep an in-use flag for the palette memory location and indicate if the routine has been given a value to the memory location by dynamic palette loading. Step 108 then tests only those pallet memory locations whose in-use flag is set.

If the result of step 108 is affirmative, that is, there is an existing palette memory entry that properly approximates the current source pixel value, the routine proceeds to step 11.
Go to 0. In this step, the routine places the palette memory address for that entry in the display memory location corresponding to the current pixel. According to the present invention, the routine then updates the last used entry for the pallet memory location. That is, the last used entry for the palette memory location is set to a value indicating that the palette memory location was used at the point in the image that the main pointer is currently satisfied. As will be seen below, this prevents a dynamic palette loading operation from changing the contents of the palette memory location before it can be used to display the pixel at which it is currently encoded.

The routine then proceeds to step 114. At this step, the routine determines if all pixels of the image have been examined. If so, of course the routine stops. On the other hand, the routine advances the main pointer to identify the next pixel and, as block 116 indicates,
Resume at step 116. If step 118 determines that the existing palette memory does not approximate the current source pixel value, the routine is represented by block 118 and is shown in more detail in FIGS.
Go to the TRY subroutine. The purpose of this subroutine is
It consists in trying to find the DPL position preceding the current pixel, so that the DPL instruction is used to load the palette memory with a content that appropriately approximates the intended value for the current image. .

As mentioned above, the dynamic palette loading routine uses the DPL pointer as well as the main pointer. The DPL pointer usually specifies a position somewhere behind the position defined by the main pointer. The DPL pointer then displays the position within the image data. This location is the location where the routine next resumes searching for a DPL location, that is, a location where a normal object codeword can be replaced with a DPL instruction. Step 120 defines whether the source word identified by the DPL pointer is followed by a sufficiently long sequence of the same or, in some embodiments, approximately equal source words. The reason for this is, of course, that the display system
Because it displays the same series of pixel values at a position in the display, for this display the display memory contains DPL instructions. Therefore, the series of pixel words that satisfy the criteria of step 120 are potentially DPL locations.

Initially, the output of step 120 is always yes.
s. The first two lines were all set to black.
And all pixel words are the same in the first two lines. After the first two lines, however, this test usually fails by more than fifty percent and the routine proceeds to step 122. In this step, the routine increments the DPL pointer. Next, in step 124, the DPL
Determines whether the pixel sequence satisfied by the pointer still precedes the pixel specified by the main pointer. If no longer preceded, the instruction placed at that location cannot change the palette memory contents, affecting the display of the pixel specified by the memory pointer.

Since the DPL pointer does not meet the criteria made by step 124, the routine must therefore accept the closest existing palette value. Then, the routine proceeds to step 126. In step 126, the routine identifies a palette memory location whose content is a good approximation of the intended pixel value, but the approximation does not meet the acceptance criteria initially made by step 108. After step 127, where the routine updates the last used entry of the selected palette location to indicate that it was last used at the current pixel, the routine adopts the selected palette memory address as the object word for the current pixel. , Put it in the appropriate display memory location. Block 12
8 represents this step. The subroutine returns to the control of the main loop of FIG. 5 prior to the next pixel.

If the DPL pointer does not meet the criteria of step 124, then the DPL instruction at the pixel sequence specified by the DPL pointer may affect the current pixel. Therefore, the routine returns to step 120. In this routine, the routine searches again for the same series of pixels. The length of the sequence retrieved by the routine can be either 5 or 6 in the described embodiment. If there is a DPL instruction sequence immediately before the sequence specified by the DPL pointer, the pixel value displayed during the second DPL instruction is the first pixel value.
The same as displayed during the command. And only five identical values are needed to use the second DPL instruction. Normally six are used.

In either case, if the sequence of source pixel words passes the test 120, the routine determines that the palette memory contents are D at the location specified by step 120.
The PL instruction initiates the process of determining if it can be safely changed. Step 130 represents the first check of the last-use table having an entry for the first palette memory undergoing dynamic palette loading.

In principle, all 223 palette memory locations that are actually employed as a look-up table in the described embodiment are capable of undergoing dynamic palette loading. In practice, however, a first palette memory location is reserved for a fixed value representing black, which value is prevented from being changed by dynamic palette loading. While this is a benefit in our construction due to our particular initialization approach, it is clear that it is not necessary to practice the invention. To prevent any changes in the first black-containing position, the "first" palette table entry referred to in step 130 of our discussion is actually the second position, but is subject to dynamic palette loading. The first position.

In step 132, the routine looks up the last use table entry to see if the corresponding palette memory location needs to be used between the DPL and the image location identified by the main pointer. decide. That is, the purpose of step 132 should be to ensure that the intermediate object word assignment does not change the underlying palette entry. A positive test 122 result indicates that there is no use for the dependent pallet memory location. The routine then proceeds to step 134. In this step 134, the routine sets the table of contents entry for the palette memory location to the pixel value intended for display at the current pixel location, ie, the pixel identified by the main pointer with the object word the routine is currently determining. change. This changes the last used position relative to the palette memory position. Then step 134 also updates the last used field at that location to indicate that the last used so far occurs at the current pixel.

The routine then proceeds to step 136. In this step, the routine places the appropriate dynamic palette loading instruction in a series of display memory locations. This position starts from the position identified by the DPL pointer. This position can therefore no longer be used for another dynamic pallet loading instruction. So the routine advances the DPL pointer to the end of the instruction sequence. Here, the next search for the DPL position is started. Block 138 displays this step. The routine then proceeds to step 128. In this step 128, the routine writes the address of the palette location to the display memory location specified by the main pointer. As before, control returns to the main loop of FIG.

Returning to step 132. The negative result of the test of step 132 is that the tested pallet position is already
Indicates that it has already been selected for use in the display pixel following the identified DPL location. The contents of the pallet position therefore cannot change at the selected DPL position, so the routine proceeds to step 140. This step 140 determines if all of the palette memory locations have already been tested. If not, the routine then moves to the last used entry for the palette memory location and repeats step 132, as block 142 indicates. This continues until it finds a palette memory location that is not currently selected for use beyond the identified DPL location. At this DPL position, the pointer proceeds to step 134 and continues as before.

However, if the routine spends a palette memory location without finding a location that is safely changed by an instruction at the indicated DPL location, test 1
40 result is positive and the routine has identified DP
Reject the L position and find another position in the image. In particular, as block 144 shows, the last used entry for all of the routine palette memory locations is searched to find the first encountered entry and the DPL pointer is
Increment to a value that identifies the pixel immediately after the end use indicated by this first end use entry. In other words, the routine identifies the most recently used palette memory location and indicates that the search for the DPL location should begin immediately after the last use of the palette memory location. The search for the DPL location then begins at step 120.

Some observations are in the order associated with step 126. In this step, the routine selects the closest Hallet entry, even if the palette entry is unable to approximate the intended value within the tolerances allowed in step 108. Obviously, it is preferable that the routine never reach this step. This is because the routine indicates that it has deviated from the fidelity criteria that would otherwise be used.

The number of times a routine has to resort to this approach can usually be reduced by reducing the general fidelity requirement, ie by losing the tolerances made by step 120. This is D
Reduce the rate at which PL positions are used. Step 126
The use of can be similarly reduced by losing the "equivalent" requirement of step 120 or by increasing the number of DPL positions by implanting a single color side edge in the image. However, images differ in the number of colors they require, and the fidelity required in one image may not be necessary in another. Although the routine must use step 126, it refines the routines described so far to reduce the number of events that need to keep the general fidelity level to the original image as high as possible. The tolerance can vary according to the image adopted.

One way to do this is simple, it makes it possible to run the encoder, display the resulting image to the user, and adjust the tolerances made by step 102 or step 120 or both. It also makes it possible to add borders and adjust this width to optimize the substantial quality of the image. Provision should be made to automatically adjust routine parameters to reduce the degree of human intervention required.

For example, the retrieval of the DPL position which is part of the palette entry allocation routine of FIGS. 6 and 7 in steps 120 to 124 is performed in steps 106 and 108.
It can be inserted in the main loop of FIG. This makes the index between the next available DPL position and the distance to the current pixel being tested available in the approximation test of step 108. The tolerance can thus be made dependent on the distance. The tolerance is zero for large distances and increases with decreasing distance. This distance value can be employed to change the stringency of the DPL position criteria used in step 120.

Of course, this is only the roughest approach to estimating the optimum acceptance level. This level considers only the distance to the first DPL position, not the number of DPL positions within this distance. Another improvement may be adopted with only a small adoption of this general approach. In the present test displayed by the loop with steps 120, 122 and 124, the routine exits the loop when the DPL position is found. According to this further refinement, it keeps track of where the first DPL position is located, keeps searching for another DPL position and records the total number of positions of this DPL until the position identified by the main pointer is reached. Therefore the tolerance depends on the number of intermediate DPL positions and not on the distance between the first available DSP position and the current pixel position. Yet another improvement can be applied to either of the previous approaches, but this would be to use a palette memory location whose contents can be safely changed rather than all of the unused DPL locations.
Only consider position. That is, instead of steps 120-124, the sequence between steps 106 and 108 is steps 120-124, 130, 132 and 1.
40-144.

Automatic side edge adjustment can be provided by increasing the edge size and starting over whenever the test result of step 124 is negative, ie whenever the routine exhausts the DPL position. it can. We believe these improvements result in improved functionality, but we have only actually used the simpler approach described so far.

It is clear that the dynamic pallet loading method can be applied not only to mixed run systems, but also to systems that are strictly encoded on a pixel-by-pixel basis. The contents of the description of FIG. 5 to FIG.
It relates to encoding source words that were not adopted for encoding as a mix. If the system is a mixed run system, however, the test of step 106 is required, as described above, and the result is sometimes positive, indicating that the current pixel should be encoded as a mix. In these cases, the routine proceeds to step 146, where the routine calculates the mix value to be encoded for the current pixel.

This calculation is based on the palette value now assigned to the endword for the run that is part of the pixel of interest. It is recalled that the mix run encoding routine put two "end" values in the run at the beginning of the run mix value. Then, step 146 derives the palette value assigned to the last two non-mixed words.

Many approaches are possible to calculate an acceptable mix. The difference between each red, green and blue component of the end value is calculated to determine which difference is the largest and the mix is calculated according to the line segment that yields this maximum difference. For example, the difference between the red components of the end value | x _LR −
Consider whether x _RL | is greater than the difference between the corresponding green and blue. Then, mix the source value for the pixel is X _S is 32 in _{(x SR -x RL) / (} x RR -x LR) step 148, the routine in the display memory location for the current pixel code for this mix Introduce. The routine proceeds to the next pixel. The process ends when the main loop of FIG. 4 operates on all pixels.

It is clear from the above description that images with natural or other tints can be used to increase fidelity so that they can be displayed in a system with a smaller palette memory size,
The present invention can achieve a significant increase in fidelity without the use of mixed run encoding. The present invention thus represents a major advance in the art.

[Brief description of drawings]

FIG. 1 is a block diagram of an image display system for explaining the present invention,

FIG. 2 is a flow diagram of a technique employed by the system for converting image data from a pixel-by-pixel code to a mixed run format,

FIG. 3 is a continuation of FIG.

FIG. 4 shows an exemplary sequence of mixed run code pixel words including dynamic palette loading code,

FIG. 5 is a flow diagram of a main loop of a routine employed by the system to incorporate palette memory addresses and dynamic palette loading instructions into display memory data.

FIG. 6 forms a flowchart of a subroutine called by the routine of FIG.

FIG. 7 is a continuation of FIG.

[Explanation of symbols]

 12 Echoda 14 Source 16 Display Memory 18 Display Mechanism 36 Palette Memory 37 Dynamic Palette Loader 38 Decoder

 ─────────────────────────────────────────────────── —————————————————————————————————— Inventor Gary J. Flatterola New York, Hampshire 03054 Merrimack London Court 28

Claims (2)

[Claims] 1. A palette memory having a screen consisting of a screen location corresponding to each image pixel and including, corresponding to each image element, a palette memory location having a pixel value and a palette memory address associated therewith. A display mechanism having a pixel input corresponding to each pixel of the image, the pixel input referring to a palette memory address. , And the other pixel word represents a DPL instruction that directs a change to the palette memory contents, which scans a position on the screen with a screen sequence at which pixel value is determined from the contents of the palette memory position corresponding to an image pixel. And a pixel word referenced by a palette memory address corresponds to the image pixel,
And a display mechanism for changing the palette memory contents in accordance with the DPL command, B) corresponding to each screen position, the screen position corresponding to the corresponding pixel, operative to store the corresponding object word in the display memory position, and corresponding A display memory circuit that scans the display memory location with a sequence that tracks the scan of the screen location, generates a display of the display memory output for the contents of the scanned memory location, and provides the display memory output to the display mechanism as a display input; ) Configured to receive a source signal comprising a source pixel word associated with a display image pixel and converting the source pixel word into a corresponding object pixel word, some of the corresponding object image words displaying a DPL command. Image values by referencing palette memory addresses, others The contents of the palette memory address displayed and resulting from the DPL instruction approximate the value displayed by the corresponding source word and comprise an encoder which stores the resulting object pixel at the corresponding display memory location. Device for displaying images displayed by signals. 2. The encoder includes: A) checking the source pixel value in a screen sequence; and B) the source signal for each of a plurality of pixels for displaying the pixel value, the contents of the palette memory location being present. Determines whether to approximate the display pixel value within a predetermined tolerance,
When approximated, the index of the address of the palette memory location is stored in the display memory location corresponding to that pixel, and C) the content is most recently used for each of the plurality of palette memory locations through the test equipment. Hold the last used value to display the position in the image, and D) the contents where the corresponding palette memory position does not exist,
For a pixel that approximates the pixel value displayed by the corresponding source pixel word, the source pixel word corresponding to the previous pixel in the screen sequence is the corresponding display position as the available DPL position according to a predetermined criterion. Is appropriate and stores the instruction in the available DPL location to change the contents of the palette memory location so that the last used value for the palette memory location is the previous available DPL location. 2. The apparatus of claim 1 including means for displaying the image location to approximate the pixel value represented by the corresponding source pixel word.