GB2282514A - Anti-aliasing method - Google Patents

Description

h VIDEO SIGNAL PROCESSORS 2282514 The present invention relates to video

signal processors, for example video character generators.

In television broadcast studios, character generators are used for providing character signals which are overlaid on to a television signal for displaying an image on a television receiver. To display high quality text, an anti-aliasing technique is used to filter digitally the signal representing each character before use. However, this can result in a slower processing speed than may be desired in some applications.

According to the present invention, there is provided a video signal processor in which signals representing images are subjected to antialiasing before display, the processor including means for 1) subjecting such signals to a first, relatively slow anti-aliasing process to produce relatively high quality images and 2) subjecting such signals to a second, relatively fast anti-aliasing process to produce relatively lower quality images; and means for selecting between the first and second processes.

The video signal processor is preferably a video character generator, and the signals representing images preferably represent characters or other graphical features.

The present invention will now be described, by way of example, with reference to the single figure of the accompanying drawing which is a block diagram of an example of a video character generator using the present invention.

Referring to the drawing, this is a block diagram of a video character generator of the kind made by Aston Electronic Designs Limited, of 123/127 Deepcut Bridge Road, Deepcut, Camberley, Surrey GU16 6SD, England, and called "MOTIF".

Images for display are held in a dual framestore arrangement 1 comprising a pair of 32 bit framestores A and B, which allow on-air cutting and ef fects to be carried out between two pages. The framestore arrangement is designed to present both f ields of the image simultaneously to a movement interpolator 2. This interpolates horizontally by taking a weighted sum of three adjacent pixels on the current line, and vertically by taking a weighted sum of corresponding pixel values from the current line in the current field and the lines on each side (in the other f ield). By varying the weightings, the resultant output places an image in the output page to sub-pixel (horizontally) or sub-TV line (vertically) accuracy. In this manner captions can be rolled or crawled smoothly at any speed.

Output from the movement interpolator 2 is passed for output to conventional digital to analogue converter (DAC) and analogue processing output stage 3 and a conventional digital output processing stage 4.

A separate 32 bit status framestore 5 holds a full colour status image for display to the operator.

h The display architecture described above is controlled by a display processor 6 which is implemented using bit slice components. The generator is controlled by a 68030 microprocessor 7 together with a maths co-processor which handles 16Mb of memory, including a fount cache, disk storage and communications with the operator and external peripherals.

In practice, the generator's mainframe occupies three units of rack space and is 530mm deep. It comprises a vertical back-to-back motherboard into which four 'Double Euro Cards' 280mm deep are inserted at the front and a a further two 'Double Euro Cards ' are inserted at the back. The four front cards comprise: the video output; framestore arrangement 1; display processor 6; and central processor. The two rear cards provide: the drivers and video/syncs input and output sockets; the parallel and serial interfaces; network interface; and keyboard input.

The electronics unit also accommodates a power supply, two standard disk drives (one 127Mb hard disk and one 3.5" floppy disk drive), plus an optional 44Mb Syquest disk drive with removable disks, SPG/coder keyer and digital serial and parallel outputs to CCIR Recommendations 601 and 656. Two Video Grab options are also available, one capable of importing images in GBR or Y/Pb/Pr plus Key, the other in 601 parallel and serial formats.

A keyboard 8 is provided and up to ten keyboards can be daisy-chained together using standard 75 ohm video cable, to a maximum of 300m from the mainframe.

Video outputs are provided for both edit and transmission. The edit output from a keyer 9 is provided as GBR, whereas transmission can be conf igured for either GBR or Y/Pb/Pr plus key. Separate co-timed sync outputs, derived from an external sync source or from the internal sync pulse generator, are also provided for both edit and transmission.

Two parallel interfaces are provided, together with two intelligent serial interfaces which, for electronic newsroom applications, can be configured for RS232 or RS422 operation. A parallel Centronics interface can be connected to a printer for printing directories, etc. A network interface can handle up to eight nodes, or up to 64 nodes with external hardware. Based on ARCNET, the network enables individual page files, pages, images, logos and typefaces to be transferred between machines.

An optional genlock colour SPG/coder 10 and keyer 11 arrangement provides two keyed outputs (one with a protective bypass), two coded outputs, two broadcast quality mixed sync outputs (when genlocked) and a YC(S) output. An optional digital output to CCIR Recommendations 601 and 656 is available in two versions: parallel only (transmission and key); and parallel and serial. The Video Grab option is available for inputs from GBR-Y/Pb/Pr plus key sources, or for parallel/serial digital format.

Characters (or other graphical data) to be processed and/or displayed by the system are entered at the keyboard 8 or remotely from the parallel, serial or network interface are loaded into the system via an input/output unit 12. The character generator displays the characters (or graphical data) in an anti-aliased form in order to remove visual "jaggies" or "staircasing" along diagonal lines. This is done by controlling the intensity of the pixels defining the characters more precisely than just "on" or "off". Suitably the intensity of each pixel may be selected from 256 available levels of intensity.

Perceived character quality can be judged primarily by two criteria: the "smoothness" of all lines and curves in the character (eg lack of staircasing), and the visual stability of the character (eg lack of interlace flicker). Therefore, the generator can perform anti aliasing between individual scan lines so as to avoid visual flicker which can occur because of the interlaced refresh process by which a TV picture is generated.

The method by which anti-aliasing is performed is the main factor that controls the quality of the outputted characters. However, as described above, higher degrees of anti-aliasing require longer processing times.

A One standard anti-aliasing technique is to render the required character at a larger size than required (for example at twice the final size), but in a non-antialiased manner. This non-anti-aliased master character is then reduced to the required size by passing a two dimensional filter over it. According to the example of the present invention, the 68030 processor 7 changes the characteristics of the filter to be used by the character generator, along with the required reduction ratio from the master to the required height, so that different qualities of characters and therefore different speeds of processing can be chosen by a user. For example, the user may choose between high quality characters with a relatively slow processing speed and lower quality characters with a relatively fast processing speed.

To allow this to be done at least two processing techniques must be available to the generator. Two processing techniques which require different processing speeds and produce characters of different qualities, will now be described as examples.

In a first, relatively slow technique, a two dimensional Gaussian filter is used to reduce the image and produce a high quality character. Assuming a reduction ratio R in both height and width from the master character to the anti-aliased character, a two dimensional Gaussian filter of size N pixels (wide) by M lines (high) is used to produce a weighted sum over an area of the master character. The value of this weighted sum at a particular point is used to give the intensity of the relevant pixel in the anti-aliased character. The filter is applied every R pixels and R lines across the master character to produce the required anti-aliased character size.

More specifically, the weighted sum W for a pixel at point (x,y) in the anti-aliased character is calculated as follows:

m N '2 Wv v F(.,jyj . M(p,,.iRY., (Xly) m N j) 1 2 where:

F(,.,) is the filter weight at offset (x, y) (note that (0,0) is the centre of the filter); and M(x.Y) is the pixel value, or intensity, in the master character at location (x,y). M(,Y) can only have the value If If 0 or 1, indicating "off" or on.

Typically the following constraints will apply:

N, M a: R to ensure all master pixels are sampled; and M k:N to help reduce interlace flicker.

The Gaussian filter weight F at z pixels and y lines from the centre of the filter can be calculated from:

F(,v) = k e -(,nd)2 where: k is a scaling constant used to control total filter weight; e is the natural logarithm base; and d is the distance, scaled by a constant c, from the centre of the filter to the position at x pixels and y lines from the centre of the filter, given by:

i d = c Fx2+y2 Different character qualities and different processing speeds may be produced by varying the constants R, M, N, k and c above.

A second relatively faster technique (typically providing reduced quality) is a variant of the f irst technique and uses a f lat f ilter rather than a Gaussian f ilter. In this case every filter weight has the same constant value f, that is:

F(,,J) = f In this technique, the values for M and N will typically be smaller than the equivalent values using the first technique.

The operator can select, for example by using the keyboard 8, between these techniques or, optionally, between further techniques to achieve a desired balance between processing speed and character quality.

Claims (11)

1. A video signal processor in which signals representing images are subjected to anti-aliasing before display, the processor including means for 1) subjecting such signals to a first, relatively slow anti-aliasing process to produce relatively high quality images for display and 2) subjecting such signals to a second, relatively fast anti-aliasing process to produce relatively lower quality images for display; and means for selecting between the first and second processes.

2. A video signal processor as claimed in claim 1, wherein the first antialiasing process includes the steps of receiving data defining the images at a resolution greater than that at which they are to be displayed, and processing the received data to generate the images for display.

3. A video signal processor as claimed in claim 2, wherein the first antialiasing process includes the step of applying a first filter to the received data.

4. A video signal processor as claimed in claim 3, wherein the first anti-aliasing process includes the step of determining the intensity of a pixel of an image for display by calculating the sum of the products of 1) the intensity of each relevant location in the received data and 2) the value of a first filter function appropriate to that location.

5. A video signal processor as claimed in claim 4, wherein the first filter function is a Gaussian filter function.

6. A video signal processor as claimed in any preceding claim, wherein the second anti-aliasing process includes the steps of receiving data defining the images at a resolution greater than that at which they are to be 1, 1 displayed, and processing the received data to generate the images for display.

7. A video signal processor as claimed in claim 6, wherein the second anti-aliasing process includes the step of applying a second filter to the received data.

8. A video signal processor as claimed in claim 7, wherein the second anti-aliasing process includes the step of determining the intensity of a pixel of an image for display by calculating the sum of the products of 1) the intensity of each relevant location in the received data and 2) the value of a f ilter function appropriate to that location.

9. A video signal processor as claimed in claim 8, wherein the filter function is of a constant value for all locations.

10. A video signal processor as claimed in any preceding claim, wherein the video signal processor is a video character generator, and the signals representing images include representations of characters or other graphical features.

11. A video signal processor substantially as herein described with reference to the accompanying drawing.