AU8936698A - Media pipeline with multichannel video processing and playback - Google Patents

Description

i.J d~t Regulation 3.2

AUSTRALIA

Patents Act 1990 COMPLETE SPECIFICATION FOR A STANDARD PATENT

(ORIGINAL)

I

r r r o r Name of Applicant: Actual Inventors: Avid Technology, Inc., of Metropolitan Technology Park, One Park West. Tewksbury, Massachusetts 01876, United States of America KURTZ, Jeffrey CACCIATORE, Ray ZAWOJSKI, Peter PETERS, Eric, C.

WALSH, John, Jr.

DAVIES COLLISON CAVE, Patent Attorneys, of 1 Little Collins Street, Melbourne, Victoria 3000, Australia "Media pipeline with multichannel video processing and playback" Address for Service: Invention Title: The following statement is a full description of this invention, including the best method of performing it known to us: 1is 4 j i:: ~B:1 -lA- MEDIA PIPELINE WITH ULTICRANNEL VIDEO PROCESSING AND PLAYBACK Technology for manipulating digital video has progressed to the point where it can be readily processed and handled on computers. For example, the Avid/i Media Composer, available from Avid Technology, Inc. of Tewksbury, Massachusetts, is a system wherein digital video can be readily captured, edited, and displayed for various purposes, such as broadcast television and film and video program post-production.

The Avid/i Media Composer uses a media pipeline to provide real-time digital video output on a computer display. This media pipeline 30 is shown in Fig. 1 and is described in more detail in U.S. Patent No. 5,045,940, issued September 3, 1991. In this media pipeline 30, a permanent storage 40 stores sequences of digital still images which represent digital video and are played back at a rate which provides the appearance of video. The sequences of digital still images do not include any frame synchronization or other type of timing information which are typically found in television signals. The still images also typically are stored in compressed form. The stored sequences are accessed and placed in a data buffer 42 from where they are provided to a compression/decompression system 44. The output of the compression/decompression system 44 is applied to a frame buffer 46 which converts the still image to a typical video S'signal which is then applied to an input/output unit 48.

SEach of the systems 40, 42, 44, 46, and 48 in this media pipeline 30 operate bi-directionally. That is, the output process discussed above can be reversed and video signals can be input via input/output unit 48 to the frame buffer 46 where they are converted to a sequence of digital still images. The images in the sequence are compressed by I compression/decompression system 44, stored in data buffer 42 and then transferred to the permanent storage -1 -RI- ~----LI-UUIIIIU-n l-1C P 'O'ERiCMI.'AVI) DIV- -2- Although the media pipeline provides many advantages for digital video, including enabling broadcast and editing from the stored digital images in a computer system, this media pipeline is not able to provide real-time digital video effects including complex arbitrary threedimensional effects, simple, two-dimensional effects such as resizing, x-y translation, rotating, layering (an appearance of picture-in-picture), and finally simple effects such as dissolves, wipes, fades and luma and/or chroma keying. In order to perceive such effects on the computer, the effect generally is first generated (not in real time), then digitized and stored if generated on a tape and finally played back.

As additional background, most digital video effect systems operate on and mix analog video signals together. Such a system is shown, for example, in US Patent No. 4,698,682, to Brian Astle.

The invention improves over the prior art by providing a media pipeline with two channels for processing sequences of digital still images. A blender is provided so as to enable simple effects on these two streams of video data such as dissolves, wipes and chroma keys.

15 Complex arbitrary three-dimensional effects and other effects may also be provided using an external interface.

Thus, a system for processing sequences of digital still images to provide real-time digital video effects includes first and second channels for communicating first and second sequences of digital still images at a rate for simulating video. A controller directs still images 20 to one of the first and second channels. A blender, having a first input connected to the first channel, a second input connected to the second channel, and an output, provides a combination of the first and second sequences of digital still images at a rate for simulating video.

The invention is described in greater detail hereinafter, by way of example only, with reference to the accompanying drawings, in which: Fig. 1 is a block diagram of a media pipeline as is used in the prior art; Fig. 2 is a block diagram of a modified media pipeline in accordance with an embodiment of the present invention; Fig. 3 is a more detailed block diagram of a modified compression/decompression subsystem of the media pipeline in accordance with an embodiment of the present invention; Fig. 4 is a block diagram of a modified media pipeline in accordance with an embodiment of the present invention to provide real time digital video effects; Fig. 5 is a block diagram of the a generator for box wipes; I" ,t*I'OIL 'JC L+IV, 'M -3- Fig. 6 is a flow chart describing the operation of a state machine for each scan line in a frame for a box wipe; Fig. 7 is a flow chart describing the operation of a state machine for each pixel in a scan line for a box wipe; and Fig. 8 is a diagram showing how a is determined for different regions of an image for a box wipe.

The present invention will be more completely understood through the following detailed description which should be read in conjunction with the attached drawing in which similar reference numbers indicate similar structures. All references cited herein, including pending patent applications, are hereby expressly incorporated by reference.

A media pipeline 35 with two channels of digital video for providing effects will now be described in connection with Fig. 2. The media pipeline 30 shown in Fig. 1 is modified to include a compression/decompression (CODEC) unit 58 which is a modification of compression/decompression system 44 of Fig. 1. The CODEC unit 58 has two CODEC 15 channels 50 and 52. One is used for compression and decompression, for both recording and playback, while the other is used only for playback. The outputs of these channels are fed to a blender 54 which combines them channels .are fed to endenot necessary to use according to the desired effect. t is not necessar to ue compressed data; however, compression is preferable to reduce compressed data; howevhs preson/decompression unit 58 storage requirements. This compession/ecopresson unt 58 is described in more detail in British provisional specification 9307894.7, filed April 16, 1993, under

U.S.

foreign filing license 504287 granted April 13, 1993.

This CODEC unit 58 will now be described in more detail in connection with Fig. 3. In this figure, a control unit controls two channels of coder/decoders. The modification to L the media pipeline.

30 is made by assigning, in the control unit 60, different sections of the compressed data buffer 42 68 is based on the assignment of a channel to a selected Ssequence of digital still images.

C

Each channel has a separate CODEC, either a first

CODEC

66 or a second CODEC 68. The CODEC typically use the Joint Photographic Expert Group (JPEG) proposed standard for still image compression- Such CODECs are commercially available, such as the CL550 available from C-Cube of Milpitas, California. Each CODEC has a respective first-in, first-out (FIFO) memory elements 62 and 64. The FIFO memory elements :62 and 4 feed d respectively to the CODECs 66 and 68 of which Sthe outputs are applied to field buffers 70 and 72, which are h-e o t e a ppThesee to ha lnels m also preferably FIFO memory elements. These two channels may be blended using a blender 74 which is controlled by an addressing and alpha information unit 76, as will be described in more detail below. The blender 74 and alpha and addressing information unit 76 are preferably implemented using a field reprogrammable gate array such as the XC3090 manufactured by XiLinx.

Alternatively, a first output sequence may be provided by output A from the FIFO 70 for CODEC 66, and a second output sequence may then be provided by the output of the blender, when no blending is performed, as the output B. Thus, FIFO and blender 74 act as first and second sources of sequences of digital still images. The outputs A and B may be applied to a digital video effects system 59 as shown in Fig. 4. This embodiment is useful for providing arbitrary three-dimensional video effects, as are described in U.S.

patent application entitled Media Pipeline with Mechanism for Real-Time Addition of Digital Video Effects filed March 18, 1994 by Harry Der et al., and assigned to Avid Technology, Inc. of Tewksbury, Massachusetts.

More complex, two-dimensional effects can also be made using techniques known in the art, including X-Y translation, rotation and scaling. An additional effects board, similar to that for the three-dimensional, arbitrary effects, can be provided so as to perform an operation on a single stream.

To provide this operation, the output A as shown in Fig. 4 is applied to such an effects generator, the output of which would be applied to the input of the blender originally designed to receive channel A. When this capability is provided, the digital effects typically produce an output using the YUV data format with four bits for each of the Y, J and V parameters In contrast, the normal data format for channels A and B is 4:2:2. Thus, in this instance, the blender 76 should be designed so as to optionally process channel A for either 4:4:4 format or 4:2:2 format, according to whether such digital effects are being provided.

Provision of simpler video effects in real time, such as box wipes and chrm.-a and luma keys, using blender 74 and alpha and addressing information 76, will now be described.

Blending of two s7 eams (A and B) of video typically involves the application of the function mA (1-a)B to the streams of video information, where a is a value which can vary from pixel to pixel in an image, and where A and B, at any given point in time, are pixels in corresponding frames in the two streams (A and B) of video. Each effect is thus applied to one frame from each of the two streams. (One normally does not perform an effect on only a fraction of a frame). Given a, at any point in time and the addresses for pixels A and B, the output image can be generated. The blender 74 which performs this operation be determining the result of the combination oA (l-a)B can be implemented using standard digital hardware design techniques.

Preferably, a field reprogrammable gate array is used to implement the function (A-B)a

B.

The value of a applied to two pixels is dependent upon the kind of effect to be provided. For example, dissolves uses the same a for for all pixels in one frame. a is

S

gradually decreased for subsequent frames in the dissolve.

The pair of pixels to be combined is also dependent upon the kind of effect to be provided. The indication of which pair of pixels to use is called the addressing information. For each kind of effect to be provided, a state machine and state variables can be defined for processing one frame of output video. The two general types of effects are chroma and/or luma keys and box wipes, which include dissolves and fades.

For example, in order to implement chroma and/or luma :keying, two thrtshold values Dl and D2 and a key point Kc are defined by the user for each parameter of a pixel. These effects are typically applied to the YUV representation of an image. Thus, an incoming image is processed by comparing the pixel Y, U and V values to the key points and threshold values defined for Y, U, and V. In particular, using the parameter U as an example, IIKc-UI is calculated. If this value is less than D1, a is set to be the maximum possible value. If this value is greater than D2, a is set to be the minimum possible value. When the value is somewhere 4.

R -Mft*AC~UY.Z ~L I tMWX^

I

t t r r r -7between D1 and D2, a value for a is determined according to this value. In one embodiment of the invention, D1-D2 is required to be some fixed number, 16. The magnitude of this fixed number represents the desired number of a. In this embodiment, when the value of jIKc-UjI is between D1 and D2, the value I[Kc-UII-D1 is applied to a lookup table (stored in a random access, preferably rewritable, memory), which stores corresponding values of a to be used. The values of a may be any function of the input I!Kc-UII-D

I

such as a step function, a sigmoid function, a ramp or any function desired by a user. Typically, only Y or U,V are keyed and processed. One could apply keying to all of Y, U and V at once, and combine the resulting a values, for example, by using the function (1/2(au+v)

AND

ay).

Box wipes with a-border can also be provided. A box wipe is a transition between two streams defined by a rectangular shape. Within the rectangle, information from one channel is provided. Outside the rectangle, information from the other channel is provided. The transition region can be strictly defined by the border of the rectangle or a border color can be provided. The transition can be described as a linear ramp (defined by a ratio of the channels to each other). The transition is thus defined by the lower and upper limits of the ramp, the step size, and the duration. All of these parameters should be user definable. Also, the coordinates of the box should be programmable to provide a horizontal wipe, a vertical wipe, or some corner to corner wipe.

Typically, a blend is performed from the first channel to the border, from the border to the next channel, or among both channels. A state machine can readily be defined according to the variables defining the wipe so as to provide an output a value for each pair of pixels to be combined. There are three values used to define the final a. The ainit values define the initial and a, values, where mX and my are accumulated iues according to the -8state machine. In the simplest wipe, a dissolve, the initial values are held, not changed, throughout a whole frame. In the other box wipes, aX and ay may change, according to the desired wipe. In this process, the final a value is typically taken to be subject to a limiting function defined by my. That is, the final a typically is aX when mX is less than ay and typically is my when X is greater than ay.

A wipe is defined by two sets of parameters. The first set is parameters for the X direction in a frame; the second set is parameters for the Y direction, defining changes between scan lines in the effect. Both of the X and Y parameters include four groups of four parameters, each group representing an operation, including offset, control, interval, and delta information. The offset information defines where blending is to begin. In the X direction, it identifies the pixel in the scan line where the first blend Sbegins. In the Y direction, it identifies the scan line where blending is to begin. The next information is control information identifying whether further operations in the scan line, or in the frame will follow. For the X parameter, this control information is represented by two bits, wherein the first bit represents whether video is swapped between the SA and B channels. The other bit indicates whether another operation in the X direction will appear. After the control information is the interval over which the blend is to be performed. The interval either identifies the number of scan lines or a number of pixels within one scan line. Finally, the delta information represents an increment to be added to :aX or Cy for each pixel over the defined interval.

Thus, a wipe is defined by four groups of four operations in each of the X and Y directions. The first operation signifies the transition from channel A to the border; the second from the border to the second channel; the third from the second channel to the border; and the fourth from the border to the first channel. If there is no border, only two -9operations are used and the second operation indicates that there is no further operation to be performed either for the scan line or for the frame.

Given the operations defining the wipe to be performed, including the four groups of operational information for each of the X and Y parameters, a state machine can be used to determine a for each pixel in the frame. These state machines will now be described in connection with Figs. through 8.

Fig. 5 is a block diagram illustrating some structures controlled by a state machine. The state machine's operation will be described in connection with the flow charts of Figs.

6 and 7. In Fig. 5, X parameter memory 82 and Y parameter memory 80 store the operations to be performed. An address pointer is stored in registers 84 and 86 for each of these memories as X and Y address pointers. Initial X and Y delta values are also stored in registers 88 and 90. These are fed to accumulators for the X and Y values 92 and 94 via switches 96 and 98. The output of the accumulators 94 and 92 are fed to compare and switch unit 100, the output of which provides the a value in a manner to be described below in connection with Fig. 8. There is also a loop counter 102 which indicates the part of frame on which the effect is being performed. The significance of this will also be discussed further below in connection with Fig. 8. There is also a Y •position counter 104 and an X position counter 106 which are used by the control 108 which operates in accordance with the flow charts of Figs. 6 and 7.

Figs. 6 and 7 will now be described.

Upon a horizontal reset (HRST) or vertical sync (VSYNC).

as indicated at step 110, the Y accumulator 94 and Y address pointer 86 are cleared. An initial Y delta value 90 is loaded then into the accumulator 94 via switch 98 (step 112). An offset is then read from the Y parameter memory into the Y pos4-ion counter 104 (step 114). Control information is then read from the Y parameter memory 80 into the loop counter 102 (step 116).

When valid data is available to be processed, and until the Y position counter 104 is not zero, operations on a scan line are performed in step 118 as will be discussed below in connection with Fig. 7. After each scan is processed, the Y position counter 104 is decremented. When the Y position counter reaches zero, the interval is read and loaded from parameter memory 80 into the Y position counter 104 (step 120). A delta value is then read from the Y parameter memory into the Y accumulator 94 and is added to the current value therein (step 122). This value is added for each scan line until the Y position counter is then zero. Each scan line is processed in accordance with the steps described below in connection with Fig. 7. When the Y position counter 104 is zero, the control information is examined in step 124 to determine if further Y operations are to be performed. If there are further operations to be performed, processing returns to step 114. Otherwise, the system waits until another horizontal reset or vertical sync occurs.

Operations on one scan line will now be described in connection with Fig. 7. These operations begin in step 130 upon the receipt of a horizontal sync or reset. In step 132, the X accumulator 92 and X address pointer 84 are cleared and an initial delta value at 88 is then loaded into the X accumulator 92 via switch 96. An offset is then loaded from the X parameter memory 82 in step 134 into the X position counter 106. Next control information is read from the X parameter memory 82 into loop counter 102 in step 136. The X position counter is decremented when valid data is available, until the X position counter 106 is zero (step 138). The interval is then read from X parameter memory 82 into X position counter 106 in step 140. The delta value is then read from X parameter memory 82 into the X accumulator 92 and is added to the current value in the accumulator 92 until the X position counter is zero (step 142). The X address pcinte-r -ll- -11and Y address pointer are incremented along this process to identify the correct operation. If the control information indicates that more X operations are to be performed, as determined in step 144, processing returns to step 134.

Otherwise, the system waits in step 146 until another horizontal sync or reset occurs.

For each pixel in the scan line, as indicated by each decrement operation on the X position counter 106, an a value is output from the compare and switch unit 100. How this operation is provided was discussed above. Further details for more complicated box wipes in this selection will now be provided in connection with Fig. 8.

As indicated in Fig. 5, the compare and switch unit 100 receives an a value from the Y accumulator 94 and aX value from X accumulator 92 and a loop counter 102. The loop counter indicates which quadrant in a box wipe with a border is being processed. The indication of a quadrant can readily be determined by the status of the state machine and a counter. Because the X operations and Y operations are each defined by four groups of four parameters, wherein each group identifies an operation to be performed when a portion of an image, there are precisely sixteen combinations of X parameters and Y parameters, each identifying a quadrant of S: the resulting image. The a value for each quadrant has a predetermined relationship with the oX and a :values. Thus, according to the loop control 102, the appropriate selection among oX and -y can be provided.

The relationships of a to aX and My for each quadrant will now be described in connection with Fig. 8.

Fig. 8 illustrates 25 regions of a box wipe with a border and transitions between one image, a border color, and another image. There are fifteen general types of regions to be considered in this effect, each being numbered accordingly in the upper left hand corner of the box. For example, the first region 200 is labelled zero as indicated at 202. A box 204 is shown in each region identifying the source of image 'iih i

[I

-M i 17C T.S .4t.o -12data (where CHO corresponds to the channel applied to input A of the blender 76 and CH1 corresponds to the input applied to input B of the blender). The first line in box 204 indicates the a value to be provided to the blender. indicates that ax is supplied as a, as in regions 4, 7, 8 and 11. "AL" indicates that ay is supplied as the a value, as in regions 1, 2, 13 and 14. "AL:A AL" indicates that ax is provided when it is less than ay and ay is provided otherwise, as in regions 0, 3, 10, 12 and 15. "AL: =AL" indicates that ce when a is greater than oy and ax is provided, as in regions 5, 6, and 9.

Having now described a few embodiments of the invention, it should be apparent to those skilled in the art that the foregoing is merely illustrative and not limiting, having been presented by way of example only. Numerous modifications and other embodiments are within the scope of one of ordinary skill in the art and are contemplated as falling within the scope of the invention as defined by the appended claims and equivalents thereto.

Throughout this specification and the claims which follow, unless the context requires otherwise, the word "comprise", and variations such as "comprises" and "comprising", will 15 be understood to imply the inclusion of a stated integer or step or group of integers or steps S: but not the exclusion of any other integer or step or group of integers or steps.

j o* 3:

Claims (1)

13- a 13 THE CLAIMS DEFINING THE INVENTION ARE AS FOLLOWS: 1. A circuit for processing sequences of digital still images to provide real-time digital video effects comprising: a first channel for communicating a first sequence of digital still images; a second channel for communicating a second sequence of digital still images; a control unit for directing still images to one of the first and second channels: and a blender has a first input connected to the first channel, a second input connected to the second channel and an output providing a combination of the first and second sequences of digital still images; wherein each digital still image of the first sequence, the second sequence and the combination of digital still images represents a frame of video information, without synchronization pulses of a video signal, such that each sequence of digital still images appears as a moving picture when displayed at an appropriate rate. 2. The circuit of claim 1, further comprising a frame buffer having an input connected S*to the output of the blender and an output providing a.video signal. 3. The circuit of claim 1, wherein the first and second channels each include a first-in, first-out memory circuit for communicating the digital still images to the blender. 4. The circuit of claim 1, wherein the first and second channels each have an input for receiving a compressed digital still images into a first buffer memory, a decompression circuit 'having an input for receiving the compressed digital still images and an input for receiving the compressed digital still images and an output providing decompressed digital still images to a second buffer memory, wherein the second buffer memory communicates the decompressed digital still images to the blender. 4 5. The circuit of claim 4, wherein the decompression circuit is a circuit which receives and decompresses images compressed using JPEG decompression. I \)PlER lCMWAVIDl 1)IV. ,l -14- 6. The circuit of claim 1, wherein the first channel is connected to first input of a digital video effects system and the output of the blender is connected to a second input of the digital video effects system and the output of the digital video effects system is connected to a framer buffer, wherein the blender is programmable so as to pass the sequence of digital still images of the second channel to the output without blending with the sequence of digital still images of the first channel. 7. The circuit of claim 3, wherein the circuit further comprises an addressing and alpha information unit which receives an indication of an effect to be performed and provides a value to the blender for each pair of pixels to be blended according to the indicated effect, wherein the blender performs a blending function on a first pixel A from the first channel and a second pixel B from the second channel such that the output of the blender is -A (1 )B. I I :15 8. A method for blending first and second sequences of digital still images to create a third sequence of digital still images, each digital still image of the first, the second, and the third sequences of digital still images representing a frame of video information, without synchronization pulses of a video signal, such that each sequence of digital still images appears as a moving picture when displayed at an appropriate rate, the method comprising the steps of: storing in first and second buffer circuits, respectively, an image from the first and second sequences; receiving an indication of a blend to be performed on the images; sequentially selecting corresponding pairs of pixels from each of the images from the first and second sequences and selecting a blending function according to the indicated blend and a relative position of the selected pixels in the desired output image; and generating an image in a third sequence of digital still images. i~i'~ 'I P \'OPER\JCMAVID DIV I611adt 9. The steps, features, compositions and compounds disclosed herein or referred to or indicated in the specification and/or claims of this application, individually or collectively, and any and all combinations of any two or more of said steps or features. DATED this 16th day of October, 1998 AVID TECHNOLOGY, INC. By its Patent Attorneys DAVIES COLLISON CAVE 1. r r r cc r r i Ir r r r. r I a 7I