Image analysis – Image enhancement or restoration – Edge or contour enhancement
Reexamination Certificate
1998-12-23
2003-07-08
Lee, Thomas D. (Department: 2724)
Image analysis
Image enhancement or restoration
Edge or contour enhancement
C382S305000, C382S173000, C382S266000
Reexamination Certificate
active
06591020
ABSTRACT:
BACKGROUND OF THE INVENTION
The present invention relates to processing images and, in particular, to processing color images in a page description language environment. It finds particular application in conjunction with processing color images including graphical objects having at least one abutting and/or overlapping edge, and will be described with particular reference thereto. It will be appreciated, however, that the invention will also find application in processing images including text and graphical images in other arrangements, and the like.
The use of graphics in computer and communication applications is very widespread and is becoming increasingly more prevalent. A computing device often transmits a document including graphical data to a printing device using a page description language. Page description languages (e.g., the PostScript® language) include interpretive programming commands useful for implementing powerful graphics capabilities. When transmitted via a page description language, graphical data is typically first converted to a bit-map data file. Printing devices then convert the bit-map data to appropriate marks on paper.
The PostScript® page description language (“PostScript”) uses what is known as a painter's algorithm. In other words, when objects are processed using PostScript, the objects are processed in back to front order. More specifically, objects in the lowest “layer” of the image are processed first. Then, objects in successively higher layers are processed.
Objects, as commonly described in page description languages, generally fall into one of three classes. The most common objects are text. These are received by the rasterizer as records that contain character, font, size and placement information. These records are converted to bitmaps, which are used to mask the current color when combining them with the existing image. The bits in the mask that are “on” correspond to pixels in the image, which are replaced with the current color; bits that are “off” correspond to pixels that are left unchanged. Generally speaking, text must be rasterized at a high resolution. However antialiasing (described below) makes it possible to rasterize text at a moderate resolution. Color precision is of relatively little importance in text: 16-32 shades of gray is generally sufficient.
A second type of object is commonly known as a pictorial image. Such an object is received as a collection of pixel intensity values (possibly color), along with position and transformation information. The object's size and shape may be altered by the transformation (e.g., it may be rotated and sheared). Generally speaking, pictorial objects are acceptably rasterized by receiving and printing them at a relatively low (300 spots per inch (“spi”) or less) resolution. Color precision, on the other hand, is relatively important for pictorials. 100-200 shades of gray (or of each of red, green and blue) is often considered a minimum for acceptable color pictorials.
The third type of object is known as synthetic graphics. Such objects are received in the form of a series of vertices describing an outline, the interior of which is to be filled with the current color. When successive synthetic graphics objects of similar colors abut, the appearance of gradual gradations may result. Alternatively, when abutting objects exhibit sharply contrasting colors, the appearance of high contrast edges may result. It is important to provide high color precision in regions of gradual gradations, while providing the appearance of high quality, high resolution at high contrast edges.
When the image is printed, the objects are arranged one on top of another on a page. An object in a higher layer will partially, or even completely, cover an object in a lower layer. Alternatively, the two (2) objects may abut one another. When printed, the edges between the overlapping or abutting objects may appear jagged. Therefore, it is often desirable to antialias these edges.
Antialiasing provides the illusion of increased resolution without incurring the costs associated with increased memory and/or increased device resolution. Processing costs are only slightly increased by antialiasing and, in some cases, are actually decreased. Costs associated with devices capable of displaying higher resolution are also avoided since antialiasing is appropriate for displaying higher-resolution data on a lower-resolution device.
The most notable improvements provided by antialiasing occur on slightly slanted lines and edges, and near the horizontal and vertical portions of gradual curves. When printed at lower resolutions without being antialiased, these features often times appear jagged. Strokes may be of any color, and edges may separate any two (2) colors. However, the most visible jagged edges are often between colors having high luminance contrast.
The tradeoffs and benefits of antialiasing depend on the resolution achievable by the final output device. Early laser printers, despite only having a capability of printing low resolution images, in many cases produced edges that appeared smooth. As xerographic technology has advanced, laser copying has become crisper, thereby producing edges that are more cleanly defined. However, one drawback to such an improvement is an increased sharpness of the edges in the image. More specifically, the sharpened edge definition makes jagged edges caused by rasterization more visible.
One approach to removing jagged edges is to increase the resolution and addressability of the output devices even further. Hence, 1,200 spot per inch printers have been developed. Depending on the contrast of the edge (which depends on the whiteness of the paper, the blackness of the ink or toner, and the illumination), the human eye perceives jaggedness up to somewhere between 1,200 and 2,400 spots per inch (at a viewing distance of about 15 cm). However, it has been shown that with gray edges, resolutions from 300 to 600 spots per inch are sufficient. In pictorial images, these lower resolutions are sufficient, provided enough levels of gray are available. Therefore, antialiasing provides a good way to represent graphical edges so that invisible information is not stored.
There are a number of known approaches to antialiasing. The simplest and best known approach is to rasterize at a higher-than-final resolution. The final image is then computed by filtering and scaling-down the high-resolution version to a low-resolution image (i.e., the final image). This method is known as supersampling. The filtering and scaling operations consist of, at each output pixel, taking an average (possibly weighted) of the pixels in the region of the high-resolution image which maps to the region of the final image surrounding that output pixel's center.
Some advantages to this approach are that it is simple and can be implemented very rapidly on almost any system. However, this approach has at least one significant drawback over other approaches. Specifically, it may substantially increase the amount of time and memory required to antialias an image.
If the image is scaled by a factor of n (e.g., 4) (which gives n extra edge positions), nxn times as much memory is used, or at least accessed, and the processing time is substantially increased. Because memory access time (especially for large quantities of memory) has not historically decreased nearly as rapidly as processor cycle time, memory access time tends to dominate total run time. Thus, the cost of supersampling increases with resolution, in proportion to the amount of memory used.
Substantially more complex methods of antialiasing exist in which the full geometry in the region of each pixel is used to compute analytically either the weighted or unweighted average. In terms of runtime complexity, the more complex methods are better than the filtering and scaling-down method described above. Furthermore, the more complex methods theoretically give better results. However, the more complex methods are also typically more difficult to implement.
The S
Fay Sharpe Fagan Minnich & McKee LLP
Lee Thomas D.
LandOfFree
Antialiazed high-resolution frame buffer architecture does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Antialiazed high-resolution frame buffer architecture, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Antialiazed high-resolution frame buffer architecture will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3094778