US 4,754,488 AGrant
Method for Ascertaining and Filling of Bounded Areas of a Colored Raster Display
Issue Date:1988-06-28
•3 Claims
•7 Drawing Sheets
Abstract
A method for use in a bit-mapped presentation display system for ascertaining the boundary of an arbitrarily-shaped closed polygon filled with a first color-coded pel pattern and then filling the ascertained polygon with a second color-coded pel pattern. The method steps comprise identifying the first color-coded pattern; determining all unique raster runs of said identified first pattern, and creating a counterpart data representation thereof; and filling in at least a portion of the raster color-coded domain with the second pattern as controlled by the data representation.
Metadata
Assignee
- International Business Machines Corporation
Inventor
- Christopher J. Lyke
Application Information
Application Number:US 6796664
Filing Date:1984-12-07
Priority Date:1984-12-07
Art Unit:266
Classifications
IPC:
G06K 900G06F 300
Field of Search:
38234036435826;28;9;22;60;17703;747;728;734518;52622;80;81;82
Patent Drawings (7 sheets)
Description
TECHNICAL FIELD
This invention relates to a method for ascertaining the boundary of an arbitrarily-shaped closed polygon filled with a first color-coded pel pattern and for filling said ascertained polygon with a second color-coded pel pattern in an interactive raster display system.
BACKGROUND
Prior art interactive buffered raster displays include low-persistence phosphor multiple gun color cathode ray tubes (CRT's). Such a CRT is the electronic medium upon which an image is painted by way of electron beam deflection and intensity modulation of its multiple colored guns. Information, both defining the image to be painted and providing the necessary control, is obtained from a buffer intermediate the CRT display and a stored program-controlled processor. The processor executes lists of graphical orders ultimately resulting in the CRT painted image. In executing its tasks, the processor causes bit values in the form of a multibit code to be stored in a counterpart location in the buffer. The multibit code, termed a "pel", consists of an x,y position coordinate and a color number or value. As each pel is extracted from the display buffer during a horizontal scan of a raster-driven display, the bits are used to index a translate table which converts them into a larger number of bits. This larger number of bits, in turn, drives designated red, green, and blue digital-to-analog converters actually modulating the multigun CRT beam intensities.
Foley and Van Dam, "Fundamentals of Interactive Computer Graphics", Addison-Wesley Publishing Co., 1982, devote chapter 12 (pp. 479-503) to the rudiments of a raster display architecture. Also, at pp. 456-460, they describe the contemporary state of the art in the ascertaining of the boundary of arbitrarily-shaped closed polygons and the filling thereof for a bit-mapped presentation/display system such as an interactive stored processor control raster color-coded display. The preferred method described on pp. 457-458 for ascertaining the shape of the polygon to be filled involves finding the intersections of a scan line with all the edges of a polygon, sorting the intersections by an increasing x coordinate, and filling in all pels between pairs of intersections.
Agui et al, U.S. Pat. No. 4,334,274, "Method of Determining Whether or Not a Region in a Picture is Within a Closed Boundary, and an Apparatus Therefor", issued June 8, 1982, describe a method for resolving ambiguous pels or pixels bordering a black/white object by edge-detecting the object and overlaying said edge with logical positive and negative object images. This permits ascertaining ambiguous pels. Thereafter, there is involved the step of either shrinking or expanding the black or white fill in the region of any ambiguous pel in order to conform to the detected object edge.
Appel and Evangelisti, "Automatic Filling of Bounded Areas in a Raster Display", IBM Technical Disclosure Bulletin, Vol. 21, pp. 1300-1303, August 1978, show the scanning and filling of a raster object starting at an arbitrary point within the object domain and then determining pel color and filling on the fly. Appel's method of filling the pel (cell) positions constituting the polygon interior requires that a cell of arbitrary color, termed a "seed" cell, be placed within the polygon. If an empty cell is adjacent (contiguous) only to the seed cell, then said empty cell is filled. The "filled" cell in turn becomes a new "seed". If an empty cell is between a cell defining the polygon border and a "seed" cell, it is left empty. Admittedly, such a recursive procedure uses no boundary equations. However, the "identify and fill" cells on the fly appear valid only for runs of simple patterns such as black objects in white fields. Indeed, where a pattern is complex such as an rXq subarray of pels of multiple colors, then the recursion becomes effectively non-terminating because where the replacement color-coded pattern corresponds at any point to the original color-coded pattern, then Appel's system of marking previously visited locations has the effect of leaving those locations unmarked.
THE INVENTION
It is accordingly an object of this invention to devise a method wherein the bounded area of a closed polygon shape in a raster display is ascertained and filled. It is a related object that the raster polygon shape be ascertained and filled even though it exhibits a complex color pattern. It is still a further object that the method execution time be finite and that any area ambiguities be completely resolvable.
These objects are satisfied by a method for use in a bit-mapped presentation display system. The method ascertains the boundary of an arbitrarily-shaped closed polygon filled with a first color-coded pel pattern, the first pattern being embedded in a raster color-coded domain. The method further fills the ascertaining polygon with a second color-coded pattern.
The method steps comprise (a) identifying the first color-coded pattern; (b) determining all unique raster runs of said identified first pattern and creating a counterpart data representation thereof; and (c) filling in at least a portion of the raster color-coded domain with the second pattern controlled by the data representation.
The determination of unique raster runs involves the steps of (d) ascertaining connected (contiguous) regions within the domain exhibiting the first pattern and forming a location list thereof by traversing the domain as if it were a connected graph from an arbitrary point therein; and (e) at the point in the traverse where it is ascertained that a domain region is not connected, resolving the boundary by comparing elements of the first pattern in high resolution with elements in the domain local to the connected and unconnected regions and amending the location list accordingly.
The advance in this art involves the recognition that contiguous runs can be represented as a graph in which each unique raster run is a node. Thus, ascertaining the polygon area can be achieved by any traverse which encounters each node only once. The traverse may be non-recursive (parent-child) or recursive (depth first). Also, the advance involves the location listing of said ascertained connected regions, resolving boundary ambiguity by high-resolution element-by-element comparison of a referenced pattern with the domain local to the ambiguity, and then filling the region controlled by the location list.
BRIEF SUMMARY OF THE DRAWING
FIG. 1 discloses an interactive stored program-controlled display and memory with a refreshed raster display configuration.
FIG. 2 sets out a black and white coded object suitable for exhibition on a monochrome display monitor.
FIG. 3 is a typical rXq primitive pattern for a complex domain (color-coded) object.
FIGS. 4 and 5 respectively illustrate a multicolor-coded object in which the multicolored pattern has been married to the black/white coded topology.
FIGS. 6-13 illustrate the determination of the unique raster runs constituting the polygon area and the corresponding list making thereof utilizing a nonrecursive traverse (parent-child).
FIG. 14 sets out the raster run control flow of the run and list-making steps of FIGS. 6-13.
FIG. 15 shows edge resolution by way of row comparison match and pel comparison upon row mismatch.
FIG. 16 sets out the connected graph representation of raster runs in FIGS. 6-13.
FIG. 17 depicts the pseudocode for a recursive traverse of the depth first type.
FIG. 18 illustrates a typical color pattern over which prior art recursive single-color raster fill algorithms would not terminate.
DESCRIPTION OF THE PREFERRED EMBODIMENT AND INDUSTRIAL APPLICABILITY
Attributes of Interactive Raster Display Systems
Referring now to FIG. 1, there is shown an interactive raster display system of the prior art. Such a system typically includes a multigun color display monitor 5, a keyboard 3 with programmable function keys, a mix of RAM and ROM based memory 7, one or more DASD's 9 attachable through an adaptor, and a stored program-controlled microprocessor 1 interconnecting the monitor, keyboard, memory, and DASD. One configuration, among many commercially available ones upon which the method of this invention may be practiced, includes the IBM 5153 Color Monitor with an IBM 4910 Color Adaptor Card, an IBM 5160 Systems Unit, a PC keyboard, and any commercially available 51/4 inch floppy disk drive with a PC adaptor.
Referring again to FIG. 1, there is also shown some detail of the refreshed raster display portion of the system. This comprises a display buffer 13, a translate table 15, a processor 1, and a refresh controller 11. In this regard, the processor 1 constructs an image of pels and writes them into the display buffer 13 over paths 27 and 29. Furthermore, processor 1 also may access and alter the contents of translation table 15 over paths 23 and 25. Once the image is resident in buffer 13, it is normally cycled through the translate table 15 and the raster display controls (not shown) as regulated by the refresh controller 11 over path 33. The refresh controller actually accesses consecutive locations in buffer 13. The color values of the extracted pels are in turn converted by the translate table and applied to the red 17, green 19, and blue 21 guns of the CRT of the color monitor 5. The x,y coordinates associated with the extracted pel control the electron beam deflection of the CRT.
Since each display buffer 13 location defines one point-sized element of an image, a simple buffer having 1 bit per pel would suffice for a two-color (black/white) image. However, color images need a multibit code for definition. To extend the color and richness of pels and patterns without burdening the buffer, the code points are used as pointers in a larger code space. These are among the functions admirably performed by structured sets of color translate tables. This is taught in Langdon et al, U.S. Pat. No. 4,255,861, issued Sept. 30, 1980. That is, a translate table serves to minimize display refresh buffer size while maximizing the number of distinctly displayable colors.
Raster Representation of Closed Polygons
Referring now to FIG. 2, there is shown a black and white raster-coded object. That is, FIG. 2 represents a bit map such as might appear in the refresh buffer or on the display face. The entire m.times.n array of pels is termed a "domain". If an "object" is defined as the replication of a reference pattern in the domain, then several objects are included in the domain of FIG. 2. Clearly, these comprise a white chain-linked object in a black field for a q.times.r white pel reference pattern and a black band having a chain link apperture set in a white field for a black pel reference pattern. In either case, the object is closed or bounded and consists of runs of contiguous pels of like color in the raster scan line direction.
Another way of viewing a run is with reference to comparing a given row of a reference pattern with the row (raster scan line) of the domain being compared. Thus, in FIG. 2 for a given r.ltoreq.n and q.ltoreq.m, r.times.q white reference pattern, the first row consists of four white pels. If comparison were made four pels at a time, then a raster run would consist of as many repetitions of four white pels in a row in the raster scan line direction. The reference pattern also indicates that the next row beneath the first row should consist of a row of four white pels. Thus, the white pels encountered in the second row of the object would be considered contiguous, whereas the black pels in the row immediately above the first row of the object in FIG. 2 having a different color indicate a border or boundary.
Referring now to FIG. 3, there is shown a typical primitive r.times.q subarray of contiguous pels, each pel of which may assume one of three, four, or more color values. In this regard, the difficulty of ascertaining the "boundary" of an object formed from multiple contiguous arrays of color complexity is greater than that of patterns of solid color, i.e. FIG. 2. This is illustrated in FIGS. 4 and 5 where the chain-linked object is somewhat masked by being embedded or formed from the color rich patterns.
The Concept of Run and Its Graph Representation
A run is operationally defined as the color value match of q consecutive pels on the ith row of a first or reference pattern with q consecutive pels on the jth row or raster scan line of a domain. With reference to FIG. 2, each row of the reference pattern consists of pels of like color value so that for a given raster scan line of the domain, a run would exhibit contiguous pels of the same color value. However, where a reference pattern has pels of several color values, then a run would require the replication on a pel comparison match basis. In this regard, reference should be made to FIG. 18 where the reference pattern consists of a q.times.r array of a pattern of red, white, and blue pels. The top row of the reference pattern is coded as blue-white blue-white and "runs" in the domain can be found in the first, fourth, and seventh domain rows from the top. Likewise, the second reference pattern row is coded as white-red white-white. This appears without ambiguity on the second, fifth, and eighth rows of the domain. As is visible in high contrast in FIGS. 2 and 6-13, a raster scan line may have one or more runs. In FIG. 6, the leftmost pel position of a run is labeled for illustration by letters. In the preferred embodiment, the location of a run is defined by the position of the leftmost pel position.
The first method step contemplates identifying the pattern constituting the object. In the simplified illustration used with FIGS. 6-13, the pattern is a q.times.r white pel array. It should be recalled that Appel et al recursively colored neighboring cells starting with the color value of a single seed cell. This recursive coloring continued until a border cell of an object was detected. While attractively simple, Appel could not use replacement color-coded patterns more complex than a single pel for his algorithm since be could not guarantee a termination of his identify and filling of cells on the fly where the replacement color-coded pattern corresponded at any point to the original color-coded pattern. This is because Appel's system of marking previously visited locations by replacing them with new values has the effect of leaving them unmarked. Reference should be made to FIG. 18. In this domain, white cells are liberally interlaced with red and blue cells. Since Appel is monochromatic, he would either white out, red out, or blue out the domain depending upon whether his seed cell was red, white, or blue. If he tried to use a more complex replacement color-coded pattern, such as a different pattern of red, white, or blue pels, it would be non-terminating. For example, if Appel's process started on a blue cell and that cell was replaced with another blue cell from a complex replacement color-coded pattern, and the same happened in an adjacent cell, then after "filling on the fly" two adjacent cells, they would be effectively unchanged, and Appel's algorithm would cycle indefinitely between them.
Referring now to FIG. 16, there is shown a connected graph representation of raster runs of the white object depicted in FIGS. 6-13. Each run is represented by a counterpart labeled vertex. Line-to-line contiguity between runs is shown by an edge connecting a pair of vertices. This means that the determination of the area of a closed polygon by scanning contiguous runs is the same as dealing with the graph and visiting the vertices and connecting edges in a systematic fashion.
The determination of the polygon area formed by unique contiguous raster runs is equivalent to presenting such runs as a graph and then traversing the graph such that each node is visited only once. In this specification, both a recursive (depth first) and nonrecursive (parent-child) traverse are discussed.
The Recursive Traverse
In this regard, a depth first search, which is a generalization of the preorder traversal of a tree, is one important technique. A depth first search works by selecting one vertex v as a start vertex and marking v as visited. Then each unvisited vertex adjacent to v is searched in turn, using depth first search recursively. Once all vertices that can be reached from v have been visited, the search of v is complete. If some vertices remain unvisited, an unvisited vertex is selected as a new start vertex. This process is repeated until all the vertices have been visited. The name "depth first search" is apt because the technique continues searching in the forward (deeper) direction as long as possible. For example, if x is the most recently visited vertex, then a depth first search selects some unexplored edge connecting x to y. If y has been visited, the procedure looks for another unexplored edge connected to x. If y has not been visited, then the procedure marks y as visited and initiates a new search at y. After completing the search through all paths beginning at y, the search returns to x, the vertex from which y was first visited. The process of selecting unexplored edges emanating from x is then continued until all edges from x have been explored.
Illustrative of a pseudocode representation of a depth first search is shown in FIG. 17. An adjacency list can be used to represent the vertices adjacent to vertex x, and a data structure such as an array mark, whose elements are chosen from (visited, unvisited), can be used to determine whether a vertex has been previously visited.
Assume that the procedure dfs(v) of FIG. 17 is applied to the graph of FIG. 16 with the equal sign A. The procedure marks A visited and selects vertex B from the adjacency list of vertex A. Since B is unvisited, the search continues by calling dfs(B). The algorithm now marks B visited and selects the first vertex from the adjacency list for vertex B. Thus, dfs(F) is in turn invoked. The algorithm now marks F. D is in turn marked and since A has already been marked, a pattern is invoked using dfs(C), dfs(G), dfs(E), or dfs(E), dfs(G), dfs(C). At this point, the original call of dfs is complete and all the vertices have been marked.
Among the properties of the depth first search is a traversal of all connected vertices in a graph and marking a memory or list to indicate those vertices visited.
The Non-recursive Traverse
Another way of visiting all the nodes of a graph exactly once is to identify all of the nodes contiguous to a given node on an "echelon basis". A more colorful terminology would recognize that any given node or vertex may be labeled a "parent" and those adjacent to it termed its "children" or "descendents". The following illustrative example exemplifies the non-recursive (parent-child) traverse and is an alternative to the recursive traverse. The factors dictating utilization of one or the other are those ordinarily appreciated by those skilled in the art and include the amount of available internal memory. It is, for example, well known that recursive procedures may require an order of magnitude or more than non-recursive procedures.
Illustrative Example
Referring now to FIGS. 6 and 14, the method of the invention utilizes three lists: a permanent final list and a temporary parent and child list. Assuming that the first pattern is identified, and starting at an arbitrary point within an object, a graph traversing search or scan is invoked. In this regard, a conventional left scan is made until the lefthand end of a run is identified and labeled, in this case A. After identifying the coordinates of the left end of a run, a scan is made to determine the length moving from left to right. In this case, run A is nine pels long. It is then placed on the "parent list" and on the "final list".
It is next desirable to identify the contiguous runs or, in the argot of FIG. 14, identifying the "dependents" or "children". Referring now to FIG. 7, the "children of run A are the contiguous runs B, C, D, and E. Since the dependents are not on the "final list", they are therefore added both to the "final list" and the "child list".
Referring now to FIG. 8, the entries on the "child list" are moved over to the "parent list". Starting with run B, runs A and F are contiguous thereto. However, because run A is already on the "final list", it does not appear on the "child list". Consequently, only run F appears on the "child list". At this point, runs B, C, D, E, and F are added to the "final list".
Referring now to FIG. 9, the dependent (contiguous) runs associated with run C are processed. These are runs A and G. Since A is already on the "final list", it does not reappear on the "child list". Thus, only run G appears on the "child list". Note that B, C, D, E, F, and G are on the "final list".
Referring now to FIG. 10, runs F and A are the contiguous descendents of run D. Because they are already on the "final list", there is no change. The same is true with run E as a parent because A and G are also already on the "final list" shown in FIG. 11. At this point, the "child list" runs of F and G are transferred to the "parent list". There are no dependent runs since B, C, D, and E are already on the "final list".
As set out in FIG. 13, the final list now completely describes the object to be filled. The procedure terminates upon exhaustion of the "parent list".
Implementation
In the following discussion, there will be first set out the 15 steps of an implementation of the method of this invention, together with a detailed description thereof.
The steps are:
1. Initialization.
2. Travel leftward on a byte-match basis to determine run's start.
3. Travel leftward on a pel-match basis to determine run's start and see if run is already on a list.
4. Travel rightward on a byte-match basis to determine run's end.
5. Travel rightward on a pel-match basis to determine run's end.
6. Put run data on list 1 and list 2.
7. Dependent runs are now the ancestors.
8. Look up one raster line for descendent contiguous runs unless already at the top of screen.
9. Check for new runs.
10. Look for new descendent runs from the end of the current run.
11. No match, so travel right one pel.
12. Look down one raster line for new dependent runs unless already at the bottom of the screen.
13. Get the next antecedent (parent) run.
14. All of the descendents have been found.
15. All runs are written to the screen.
Detailed Description of Implementation
In step 1, two run counters are created. These are the current and previous run counters and are both set to zero. Next, the workspace buffer is divided into two equal parts where list 1 stores the screen starting byte offsets (1 word), and list 2 stores the run's byte length, the end pel offset and the start pel offset. After this, list 1 and list 2 pointers are set to the top of lists 1 and 2 respectively. The list 2 pointer for a run is equal to the list 1 pointer for the run plus half the workspace buffer length.
In step 2, the starting byte offset is set to the beginning column offset, and the length and the start pel offsets are set to zero. The cursor travels left from the beginning column offset to find the run's starting location. First, it must be ascertained whether the pattern matches to the byte boundary. If so, set the byte-match flag high.
At this point, it is necessary to begin to match the background pattern with the Regen buffer (the screen) on a byte level. Each time an entire byte matches, it is necessary to decrement the start byte offset and increment the byte length. This continues until a match fails or until the left edge of the screen is reached. This location is the starting byte offset unless modified in the ensuing step.
In step 3, since a byte match has failed, it is necessary to find the pel offset and the byte offset to the start of the run. If the byte-match flag from step 2 is greater than zero, then it is necessary to ascertain whether 3, 2, or lastly, 1 of the Regen pels matches the background pattern looking left from the start byte offset. The start pel offset is found to be 1 if the closest 3 pels match, 2 if the closest 2 pels match, and 3 if the closest pel matches from the start byte offset. If a non-zero pel offset is found, then the start byte is decremented and the byte length incremented.
If the byte-match flag from step 2 is equal to zero, a different procedure is followed. In this case, it is first necessary to ascertain whether the two closest pels match the background pattern looking left from the start byte offset. If the two closest pels do not match the background, then it is necessary to determine whether the closest pel matches. If no match is found, then the start pel position is the remainder when the beginning column offset is divided by 4. If a match is found, then the starting pel position is decremented by the number of pels that matched.
With this starting location, i.e. both byte and pel offsets, it is necessary to check list 1 first for a match, and then list 2's starting pel offset for confirmation. If the current run is already on list 1 and list 2, then go to step 10.
In step 4, travel rightward on a byte-match basis to determine the run's end requires that the end pel offset be set to 3 and the byte-match flag set to low. From the beginning column offset, travel towards the right. Determine whether the pattern matches to the byte boundary. If so, it is necessary to set the byte-match flag high and then begin to match the background pattern with the Regen buffer on a byte level and each time an entire byte matches, then increment the byte length. This continues until a match fails or the right edge of the screen is reached. At this point, the run length has been found unless modified by step 5.
In step 5, travel is rightward on a pel-match basis to determine the run's end. Since the byte-match failed, it is necessary to find the pel's end offset and the run byte's length. If the byte-match flag from step 4 is greater than zero, then see if 3, 2, or lastly, 1 Regen pel matches the background pattern looking left beginning at the position equal to the start byte plus the run byte's length plus 1. If a match is found, then it is necessary to increment the byte length. The end pel offset is found to be 2 if 3 pels match, 1 if 2 pels match, and zero if 1 pel matches.
If the byte-match flag from step 4 is equal to zero, a different procedure is followed. At this point, it is necessary to determine whether the 2 closest pels match the background pattern looking right from the beginning column offset divided by 4. If they do not match, then it is necessary to check if the cloest pel matches. The end pel offset is equal to the beginning pel offset from step 3 plus the number of pels that matched.
In step 6, it is necessary to put run data on list 1 and list 2. If the run's length in pels makes for a valid run, then the starting byte offset is placed onto list 1 and the run's byte length, and the ending and starting pel offsets onto list 2. At this point, the current run count number is incremented. Also, the current list 1 and list 2 pointers are incremented so that they point to the first empty space in their buffers. In the event that there is no more room in the buffer, then an error flag is set and a transfer made to step 15. If the previous count number is not zero, then a transfer is made to step 10.
For scanning purposes in step 7, the dependent runs are now the ancestor (parent) runs. In this regard, the descendent run list becomes the ancestor run list by setting the ancestor run list (parent) pointers into list 1 and list 2 to the start of the dependent runs in list 1 and list 2. This is accomplished by backing up the pointers from the first empty space on list 1 and list 2 by the number of dependent (children) runs. The pointers for the next generation of dependent runs continues to point to the first empty space on list 1 and list 2. Relatedly, the dependent run count number becomes the parent count number and the dependent count number is reset to zero.
In step 8, it is necessary to scan or look up one raster line for new children (dependent runs) unless the dependent scan line is already at the top of the screen. This is implemented by obtaining the starting byte, starting and ending pel, and byte length of an ancestor and determining the new starting byte if the parent were one line up. Also, the run's beginning and ending column numbers are ascertained as well as the beginning and ending byte of the entire line in the Regen buffer for one line up.
In step 9, new runs are checked. This means that if the screen pel matches the background pattern, this is treated as a new beginning point and a return is made to step 2. Otherwise, a transfer is made to step 11.
At step 10, a scan is made for new dependent runs from the end of the current dependent run. Thus, a skip is made to the end of the currently identified valid dependent (child) run, two columns are added, an update of the next dependent run's trial starting byte is performed. With respect to the latter, a check is also made to see whether this is greater than the ending column of the parent run. If so, then a transfer is made to step 12. Otherwise, a start is made from this new starting column and a transfer to step 9 performed.
In step 11, in the event of no match, travel is made right to one pel. Since the Regen buffer pattern does not match the background pattern, travel right one pel, and an update is made of the next dependent run's trial starting byte. After this, a transfer is made to step 9 unless past the end of the ancestor (parent) run. In this latter event, a transfer is made to step 12.
In step 12, a scan down of one line for new dependent runs is made unless already at the bottom of the screen. A transfer is made to step 13 if step 12 has already once been performed. From the parent starting byte, it is necessary to determine the new starting byte if that parent were one line down. Also, the run's beginning and ending column number, and the beginning and ending byte of the entire line in the Regen buffer for one line down must be recovered.
In step 13, the next ancestor (parent) run is obtained. This signifies that all the dependent (children) runs of the parent run have been identified and placed on the dependent (children) run list. The parent run is no longer needed so that the previous run count is decremented by 1 and the parent run pointer points to the next parent run in list 1 and list 2. After this, a transfer is made to step 8 until the previous run count is decremented to zero.
In step 14, all of the dependent runs have been found. Thus, all of the dependents of all of the parents have been found and put on the run list. A transfer is made to step 7 until the current run list contains zero runs.
In step 15, all runs are written to the screen. This means that all runs have been identified. Thus, beginning at the top of list 1, all of the runs can be filled with the fill pattern. The fill pattern is justified to the top lefthand corner of the screen and repeated as many times as necessary in order to fill the interior region of the ascertained shape.
The following table sets out an assembly level code sequence instructing the processor portion of an interactive raster display system in the execution of the method steps of this invention. The syntax for assembly code is from left to right:
Note, that each comment is prefixed by a semicolon (;). Also, the abbreviations "dw", "db", and "?" respectively designate define word, define byte, and reserve or allocate space.
The invention has been described at the functional, algorithmic, and assembler language levels for use in any interactive stored program-controlled raster display system. It will be apparent to those skilled in the art that various modifications and variations may be made in the method of this invention without departing from the true spirit and scope thereof.
Claims
I claim:
1. A method for use in a bit-mapped presentation/display system for ascertaining the boundary of an arbitrarily-shaped, closed polygon filled with a first color-coded q.times.r pel subarray, where q and r>1, and embedded in a raster color-coded domain of pels and for filling said ascertained polygon with a second color-coded q.times.r pel subarray, where q and r>1, comprising the steps of: (a) identifying the first color-coded pel subarray; (b) determining all unique raster runs of said identified first pel subarray and creating a counterpart data representation thereof by: (1) ascertaining connected regions within the domain exhibiting said first pel subarray and forming a location list thereof by traversing the domain from an arbitrary point therein, said traverse being selected from a set consisting of a depth-first search and a nonrecursive traverse as if each unique run were a labeled node of a connected graph; and (2) at the point in the traverse where it is ascertained that a domain region is not connected, resolving the boundary by comparing the pels of the first subarray in high resolution of at least one pel at a time with pels in the domain local to the connected and unconnected regions and amending the location list accordingly, the number of pels at a time in said high resolution comparison being less than that used for any coarse resolution; and (c) filling in at least a portion of the raster color-coded domain with the second subarray according to the data representation.
2. In a system comprising a raster display, an input device, a buffer for periodically refreshing the display, and a processor coupling said display device and buffer; the processor being responsive to the input device for altering the buffer contents; the buffer containing a raster color-coded n.times.m pel domain embedded with at least one closed polygon formed from contiguous raster runs, each run being selected from one of q rows of a first color-coded q.times.r pel subarray; each run being a rational number multiple of r pels in length, q and r>1; a method for ascertaining the polygon shape and for filling the ascertained shape with contiguous raster runs selected from q rows of a second color-coded q.times.r pel subarray; comprising the steps of: (a) identifying the first subarray; (b) ascertaining contiguous raster runs in the domain by matching selected ones of the q rows in said first q.times.r pel subarray in either a recursively descending or nonrecursive manner by comparing domain color-coded pels in a raster scan line direction with q pels of a given row in the first subarray and continuing the comparison of q pels at a time if equal, and upon mismatch, continuing the comparison on a one-pel-at-a-time basis to determine the extent of the match, a pel mismatch being indicative of the end of a run, and forming a list identifying each run by domain location and length; and (c) filling in at least a portion of the domain with contiguous raster runs formed from a rational number of multiples of r pels of a predetermined one of q rows of the second q.times.r pel subarray, the replication and positioning of the runs constituting the fill in the domain being made according to the list location and length.
3. The method according to claim 2, wherein the first and second pel subarrays each comprise a q.times.r array in which each one of q rows contains r pels of selectably distinguishable color value.
Patent Citations (10)
| Patent | Date | Inventor | Cited By |
|---|---|---|---|
| US3987412 | 1976-10-01 | Morrin II | |
| US4318121 | 1982-03-01 | Taite et al. | |
| US4334274 | 1982-06-01 | Aqui et al. | |
| US4521909 | 1985-06-01 | Wang | |
| US4528642 | 1985-07-01 | Waller | |
| US4566124 | 1986-01-01 | Yamamoto et al. | |
| US4616220 | 1986-10-01 | Grunewald et al. | |
| US4646076 | 1987-02-01 | Wiedenman et al. | |
| US4646078 | 1987-02-01 | Knierim et al. | |
| US4667306 | 1987-05-01 | Smith |
Non-Patent Literature (1)
- Appel and Evangelisti, "Automatic Filling of Bounded Areas in Raster Display", IBM Technical Disclosure Bull., vol. 21, No. 3, pp. 1300-1303.