GB2179819A - Improvements in video signals processing systems - Google Patents

Abstract

A system for manipulating video images includes parallel transfer disc stores 7 which allow video signals to be stored and recalled at the usual video rate. Once the signals have been stored on the disc stores they are then transferred in real time to a framestore 4. When read out of the framestore for processing the signals are clocked at a slower rate. This allows the processing to be achieved in non-real time and so image processing costs may be reduced. The system described may produce images of different size, shape or orientation. <IMAGE>

Description

SPECIFICATION Improvements in video signal processing systems This invention relates to video signal processing systems especially to those that manipulate images.

In television broadcasting there is often a requirement for systems which manipulate images to produce changes in shape, size, orientation etc. of that image. For example it may be desirable to create an effect where an image appears as a globe. Previous systems have taken video signals at television rate from a broadcast source or a tape, processed them in real time and then either stored the manipulated video signals on a tape or broadcast them. Where processing is done in less than real time the video signals are provided by VTR's with a slow speed capability but these can only be operated at reduced bandwidth.

One example of a real time processing system is represented in Fig. 1 and described in UK patent number 2073988. This operates by writing incoming digital video signals at video rate into framestore 1 in usual raster format but reading the signals out of the framestore in a modified format. This modified format is determined by the manipulation which is to be applied to the image and produced by providing a modified sequence of read addresses for the framestore 1. This sequence of addresses is provided by address generator 2. The addresses generated may not coincide with actual store addresses in framestore 1 and in this case the output signals are synthesised by interpolating among the video signals written in the storage locations adjacent to the addresses of the output picture points.To achieve this result signals have to be read from more than one address and subjected to interpolation to produce the output picture point within a pixel interval. The interpolated video signals are applied to display 3, after conversion to analogue signals, where the manipulated image will be seen. The process of interpolation is well known in video effect systems.

The requirement that the manipulation of the image must be at video rate restricts the type of shape changes etc. that can be achieved using this type of system. In order that the processing can be done in real time the address generator 2 must generate addresses at a rate high enough to achieve the interpolation. Where complex manipulations of the image are required the address generator must calculate addresses from complicated algorithms and conventional computers cannot do this at the required rate. Systems that can provide complex manipulation are usually very expensive.

In some circumstances it may be preferable to have a system which could receive video signals at video rate and store these signals and then, after manipulation images in non-real time, to able to view these manipulated images in real time. This type of system would allow images to be produced with very complex changes produced without being prohibitively expensive.

The object of the invention is to provide such a system without producing the changes on degraded images.

According to the invention there is provided a video processing system for image manipulation comprising, input means for video signals representing an image, disc store means, parallel transfer means for transferring a frame of video signals to said disc store means in a frame period, framestore means, transfer means for transferring said video signals to framestore means in a frame period, processing means for producing video signals representing a manipulated image, said processing means including means for generating addresses in said framestore means from which video signals must be read to produce said manipulated image, said manipulating means manipulating a frame of video signals in more than a frame period, means for displaying said manipulated video signals.

Preferably said processing means includes means for producing the manipulated video signals by interpolating video signals from a plurality of framestore addresses.

Preferably said generating means includes a computer. Further preferably said generating means includes means for storing a frame of modified addresses.

One embodiment of the invention will now be described with reference to the accompanying drawings.

Figure 1 shows a prior art system for producing manipulated image in real time.

Figure 2 shows one embodiment of the processing part of the invention.

Figure 3 illustrates interpolation of video signals.

Figure 4 shows one embodiment of the disc store part of the invention.

Figure 5 shows a second embodiment of the processing part of the invention.

As in the prior art system digital video signals representing an image are written into framestore 4 in Fig. 2 in raster format but unlike the prior art this is done at less than the usual video rate. This can be done from a disc store or from a system such as that described in co-pending application number 8506652 and shown in Fig. 4. The operation of such a system will be described in detail later, but it allows the processing to be done on full bandwidth video signals.

In the embodiment shown in Fig. 2 the computer 5 generates a set of addresses from which video signals in framestore 4 must be read in order to produce an output image which is the input image with a manipulation applied to it. Each point in the output image will have an associated generated address but these might not coincide with a store address in framestore 4.If we taken the simple case of rotation of the image as an example and consider Fig. 3 which shows the position of the output raster with respect to the input raster, the output addresses are such that information from picture points that are close in the input image is used to generated picture points which are close in the output image whereas with the more complex changes that are possible with the system of the invention output points that are adjacent may be made up of information from input points that are widely separated. The addresses for the output raster are calculated in known manner in computer 5 by applying a number of transforms and algorithms to the usual raster format until the desired effect is achieved.

Taking the address xk, yk in the output raster as shown in Fig. 3 it can be seen that this falls within a square of addresses xn, yn; xn+1, yn; xn, yn+1; xn+1, yn+1; in the input raster. The addresses generated by the computer are generated in the co-ordinate system of the input raster and include a factor which is used as an interpolation co-efficient in the interpolation of the output video signal.

The four video signals from the adjacent addresses to the output picture point to be produced are read to a 4 point interpolator 6.

This interpolator combines the four video signals read from the framestore 4 in proportions determined by the interpolation co-efficients generated by computer 5. These interpolation coefficients relate to the distance from the addresses of the four video signals in the input raster to the address of the output picture point. The computer and the four point interpolator may be of any suitable construction and examples of these are well known. Unlike the system of UK patent number 2073988 the reading out of video signals can be done in the patches of four required by the interpolator for interpolating successive output picture points.

These interpolated video signals that now represent the output image can be transferred to the disc store 7 at the slow rate used for processing and stored on a disc. Alternatively they can be written into a second framestore.

When read out in the normal addressing sequence they will appear on a display 8 as the desired image.

Although the processing works well with most manipulations when compression occurs the output signals will be unable to resolve details of short wave length in the output raster and the interpolation used to synthesise the output video signals tends to produce noticeable disturbances in the image. To overcome this problem the input signals can be filtered before applying to the framestore 4.

Fig. 5 shows one example of the processing system including a filter 10 and also an output address framestore 9. The framestore 9 shown stores a frame of addresses relating to a particular manipulation, each output point having an associated address stored in framestore 9. These addresses may be precalculated by a computer, then stored on a disc store and read to the framestore 9 prior to use. When this framestore is included, the system may again manipulate the image in near real time. Fig. 5 also shows the use of a framestore 11 which may be additional to the disc store 7 for storing manipulated pictures.

Although the processing has been described simply for luminance components of video signals it may operate with colour components, e.g., YUV in which case further framestores will be added included in the processing system.

A form of disc store for the system which allows real time video to be stored, manipulated in non-real time and then viewed in real time is shown in Fig. 4, based on our copending application number 8506652.

Incoming video signals are split into four eight bit components and written into four separate sections of the framestore 16 in real time. The system includes four disc stores, two arranged to store the luminance component and the other two the chrominance components and each of these disc stores is connected to one of the four sections of the framestore 16. Each of the four disc stores has six storage discs.

The disc stores used in this system are OEM Parallel Data Transfer Disc Drives M2350, manufactured by Fujitsu Limited and can operate in a mode where five heads in one disc store are connected so that signals can be written or read from respective tracks on discs that store in parallel. In this system twenty channels for transferring information to or from the disc stores are provided by connecting the five heads of each disc store to one of the four sections of the framestore.

When the incoming video is being transferred from the framestore to the disc stores in real time five bytes of signals are transferred in block to the five heads of each disc store and each head writes one byte of information to the disc. The reverse is true when the video signals are transferred from the disc store to the framestore for display.

In this embodiment of our invention video signals representing an image are received in real time and stored on the disc stores. When the video signals are required for processing they are transferred from the disc store to the framestore in real time. The video signals are then read from the store and processed as described previously under the control clock operating at less than the usual video rate. It is therefore possible to use a computer to provide the addresses needed, for a frame of signals in the period available for a frame which will be greater than the usual video frame period. It is to be understood that the invention is not limited to using a computer to provide addresses.

It is to be understood that the system described in UK application number 8506652 need not be used at the output end of the system.

Claims (7)

1. A video processing system comprising input means for video signals representing an image, disc store means, parallel transfer means for transferring a frame of said video signals to said disc store means in a frame period, framestore means, means for transferring a frame of signals from said disc store to said framestore in a frame period, processing means for producing video signals representing a manipulated image, said processing means including means for generating addresses in said framestore means from which video signals must be read to produce said manipulated image, said manipulating means manipulating a frame of video signals in more than a frame period, third store means for storing said manipulated video signals and display means for displaying said manipulated signals.

2. A system as in Claim 1 wherein said generating means comprises a computer.

3. A system as in Claim 1 wherein said generating means includes a store for storing a frame of modified address.

4. A system as in any of the preceding claims wherein said processing means includes means for producing the manipulated video signals by interpolating video signals from a plurality of framestore addresses.

5. A system as in Claim 1 wherein said third store means comprises a framestore.

6. A system as in Claim 5 wherein said third store means further comprises parallel transfer disc means and means to transfer said manipulated signals from said framestore to said disc store at video rate.

7. A system substantially as described herein with reference to the accompanying drawings.