CN1163531A - Image processing system with improved scaning image processor - Google Patents

Abstract

The invention relates to an improved processing system (10) for processing image signal. The system achieves maximum effective by only employing a novel scanning line image processor (12), wherein digital input is implemented as usual rate. The processor completes motion detection, motion adaptive scanning conversion, horizontal and vertical conversion, and can provide sharpness control in limited instruction space and provides the functions to four TV systems.

Description

The image processing system of improved band scan line image processor

The application be with the _ _ _ _ number be entitled as " image processing of improved employing scan line image processor " and the _ _ _ _ U.S. Patent application that number is entitled as " control of digital display definition " submits simultaneously.

The present invention relates to display system, relate in particular to the display system that picture intelligence is adopted digital processing.

Present most of display system all contains the cathode ray tube (CRT) display, this display with the scanning electron rifle with the fluorescent material of electronics " injection " to the display screen back side.Because the character of scanning electron rifle, broadcast television signal sends by interleaved form.Every frame data are divided into two, a frame that contains odd-numbered line, and another contains the frame of even number line.

Though some technology maintains the leading position in the monitor market, yet they can't adopt this Format Type.Wherein an example is exactly a spatial light modulator, but it form by matrix usually with independence selected cell, in whole addressing circuit, select by visit.The addressing circuit of these modulators is generally pressed row and column and is formed, so will to all odd-numbered lines or even number line selects then very difficult, very slow and cost is higher.

Now study certain methods this data format has been improved to the sequential scanning form, wherein can form whole frame data.This form is the pixel number by losing with each estimation of front normally, and this loses interpolation and form.Yet this interpolation process may be expensive, and calculation requirement is also higher, so the performance requirement of system and cost are improved.

The another kind of adaptability that must consider is exactly the ability that system shows different systems input data.Generally, in the CRT system, the input standard keeps a kind of analog input, and electron gun is then adjusted to new size.As for the individual unit in the array, sometimes between the pixel of image and this unit, there is man-to-man corresponding relation.For example, one has 480 row, and the system of 640 pixels of every behavior will need one to have 480 row unit, and the device of 640 unit of every behavior does not comprise the possibility of any conversion.

In this type systematic, the input data are generally in the ratio digitlization of the pixel that equals every capable size.640 pixels of every row will be by 640 samplings of every row digitlization.Therefore, processor must scalable not only can show more than one standard in the fixing array of element number, and sampling rate can with any given multiple display modes compatible.

A new development in image processing field is exactly a scan line image processor.These processors have the advantage of several respects, can be for addressing the above problem.Yet the first generation of these processors only has the extremely limited instruction space, needs to adopt several processors in a display system.The processor of new several generations has the more instruction space, and compares with last generation also to have competitiveness on price.

Therefore, people need a kind of employing scanning yoke more of new generation to resemble the system of processor, and it adopts minimum as far as possible processor, but still allows above-mentioned repertoire.

One aspect of the present invention is exactly a kind of treatment system that is used for pictorial data.This system receives data and data are formatd or takes a sample by common speed.This will allow system at different pattern of the inputs data to be formatd and handle.This system only relies on a processor to finish motion detection, Motion Adaptive scan conversion, level and vertical conversion and definition control.

An advantage of this system is that it does not need extra processor, makes its cost be significantly less than multicomputer system.

The further advantage of this system is that except other standard, it can also handle new standard, such as PALplus.

These system's further advantages are that it allows to adopt complicated FIR filter, and the existing instruction space that do not transfinite.

In order more fully to understand the present invention and further advantage, the more detailed description of do in conjunction with the accompanying drawings now thereof.Wherein:

Fig. 1 represents a block diagram that is used for the system configuration of image processing, and it only needs a scan line image processor.

Fig. 2 represents that an example can be applied to the Motion Adaptive scan rate conversion process of image data processing system.

Fig. 3 a to 3d represents the distortion to the input image signal standard.

Fig. 4 a to 4d represents the block diagram of the processing capacity that must carry out by multiple input image signal standard.

Fig. 5 represents to be applied to the vertical graphic representation that converts of a class of image data processing system.

The graphic representation that the class level of image data processing system of representing Fig. 6 to be applied to converts.

Fig. 1 represents only to adopt a complete system configuration 10 that is used for image processing of a scan line image processor (SVP) 12.Input luminance signal Y DL14 on the scene and 16 is through twice delay.These postpone to allow usually the data of access from front court and time front court.For example, if the input field at Y place is considered as 2, behind the DL14 on the scene, promptly at the A point, this will be front court, 1.Behind the DL16 on the scene, promptly at the B point, this will be time front court, 0.Field 2 wherein has the row of identical numbering with field 0, but takes from the frames of different images.The data that should add will be used for the interpolation of data, below with reference to Fig. 2 discussion.

Must be noted that when input signal Y is worked as its arrival processor 12, has been digital form generally.This can obtain from Analog signals'digitalization, runs into as the broadcast television signal of now.In addition, signal can be exactly a digital signal originally also.System shown in Figure 1 need can adopt the data of any standard on other modulators of spatial light modulator or formation displayed image.

Therefore, if data by the analog input digitlization, it must be with a kind of general sample at rates.If data are digitlization, it must convert general speed or every capable number of samples to.This describes with reference to Fig. 3 and Fig. 4.Fig. 5 and Fig. 6 will discuss the particular technology that the general rate signal that converts coincide multiple standard.All these functions can be finished by scan line image processor.The invention has the advantages that it allows these functions to be carried out by same processor, reduces hardware quantity and system cost thus.

Fig. 2 represents to be used for a kind of possible processing procedure of Motion Adaptive sequential scanning conversion.This process " is lost the field " from interleaved data fields interpolation usually, calculates the motor message between each simultaneously.Hybrid circuit 30 adopts the data from following data line, and wherein, 3 are input: the field, front court 2 of coming self-fields DL22; From the adjacent lines in time field, front court 1 of horizontal DL24 and DL26; And from the motor message k of top data channel.This top data channel is utilized input signal Y 3 and come a kind of comparison between the adjacent row of self-fields 1 at first on the spot, and it has the row of identical numbering.Obtain poor between them at the computing of label 32 or other circuit.

Formed differential signal process label is 34 horizontal low pass filter HLPF1, and obtains absolute value at ABS (absolute value) 36.Add a nonlinear function NL38.This nonlinear function forms a threshold value, to eliminate the little differential signal that produces because of noise.It also eases down to 4 with the resolution of differential signal.Comparator 44 is judged numerical value maximum in several numerical value: by the data of coming self-fields 0 of field DL40 output; By horizontal delay 42a, the data from adjacent lines to treated row of 42b and 42c output; And from 4 signals of nonlinear function 32.

Then, carry out vertical and horizontal filtering in 48 pairs of these maximums of VLPF1 46 and HLPF2 respectively.Nonlinear function 50 and label are that the similarity of 38 nonlinear function is that it has reduced signal noise.Then, formed motor message k is used for determining the output of interpolation by hybrid circuit 30.

After finishing this Motion Adaptive scan conversion, generally must convert to these data, make it to coincide with size in order to the spatial light modulator that forms displayed image.This discusses with reference to Fig. 3 a to 3d and Fig. 4 a to 4d.Fig. 3 a to 3d represents the graphic representation of four kinds of different systems.Fig. 3 a represents to be necessary for the graphic representation that the NTSC input converts and handles.

Shown in Fig. 3 a, the active pixel zone of input is 720, and sampling point is wide, 480 row are long.720 sampling rates of every row are previously discussed one routine general speed.For keeping rectangular pixels the ratio of width to height of 4: 3 ratio, must reduce being sampled to 640 sampling points from 720 sampling points of every row.Yet, selects 720 sampling points that its " integer " that becomes 9: 8 is converted and handle, rather than the conversion of any mark is handled.This conversion need not any vertical conversion.

As for the demonstration on the device, the left side of Fig. 3 a represents that device used in this example has 848 * 480 size.Its permission display standard is 640 * 480 NTSC.

Fig. 4 a represents to be used for NTSC Signal Processing flow process with block diagram.Preposition scanning (proscan) interpolation 54, motion detection 52 and definition control 56 these processing blocks keep identical in each embodiment shown in Figure 4.Brightness signal Y through motion detection 52, then, was carried out level and is converted before preposition scanning interpolation 56 at label 58 places, strengthen at label 56 places at last.Carrier chrominance signal C at first converts through level at label 60 places, carries out preposition scanning interpolation at label 54 places then.

Fig. 3 b represents a kind of standard that is called NTSC feeder connection (letterbox) is carried out similar conversion.Input active pixel zone still is that 720 sampling points are wide, but length reduces to 360 row, rather than 480 row.All the other 120 row are black surround frames of every side 60 row.By carrying out vertical simultaneously and the level conversion, such form can be converted to 848 * 480.This form has the ratio of width to height of 16: 9, and it will need image to be converted to 848 unit from 720 sampling point levels, vertically is converted to 480 cell row from 360 row.The conversion factor level is 5: 6 (720/848), vertically is 3: 4 (360/480).The horizontal black surround frame of seeing in the input cuts off from final image.

Fig. 4 b represents the processing capacity of NTSC feeder connection.5: 6 levels that luminance signal is passed through motion detection 52, preposition scanning interpolation 54 and label 62 places once more convert.Carrier chrominance signal then, enters vertical conversion with luminance signal and handles 68 through the levels conversions in 5: 6 at label 64 places, the preposition scanning at label 54 places.At last, luminance signal is strengthened at label 56 places.

Similar processing can be used for the pal mode signal.Shown in Fig. 3 c, the input data have 720 * 576 size.Yet,, only adopt 548 row in 576 row for this display application.It dwindles in 9: 8 ratio and vertical direction horizontal direction in 8: 7 ratio.The block diagram of this processing procedure is shown in Fig. 4 c.Same processing procedure can substitute the vertical conversions in 8: 7 that colourity and brightness add with label 70 places.

Aspect pal mode, be PALplus than new progress relatively.The something in common of PALplus and NTSC feeder connection is that it remains on picture intelligence the ratio of width to height of 16: 9.Yet the information that need to produce 16: 9 images is compressed in the ratio of width to height scope of 4: 3 by level.From Fig. 3 d as seen, this image must carry out level and vertical conversion, and some row must be sheared.On the level, these data must be listed as 864 row from 720 with 5: 6 ratiometric conversion.Then, must cut off 16 row to be fit to 848 row of this device.On vertical, 548 row must taper to 480 row with it by 8: 7 ratiometric conversion.Adopt the block diagram of this processing procedure to be shown in Fig. 4 d, it is used to handle colourity and luminance signal.

The filter that is used to convert may become very complicated and need become hundred instructions.Yet, since this processing design this instruction sequences in order and carefully, by adopting the processor clock of 43 megahertzes, SVP more of new generation allows up to 2730 instructions.It allows to adopt very advanced and accurate filter to convert, as shown in Figure 5.

As shown be an example, the vertical conversion coefficient when doing conversion in 3: 4 in order to discuss.In this 3: 4 conversion processes, generate 4 row with 3 row.For example, with row X2, X3 and X4 generate Y0, Y1, Y2 and Y3.By the dot product of coefficient shown in the application, can find output row Y0-Y4 with input data line X0-X7.For example, (AB) * (XY)=AX+BY.Index number is represented from this auxiliary-filter generation coefficient.For example, Y01 has coefficient-3/512,13/512, and 492/512,13/512 and-3/512.These are all from auxiliary-filter 1 generation.Filter is by 1,2,3,0,1 ... or the like order apply.And notice that this coefficient also begins repetition when used auxiliary-filter repeats.

Here used filter is finite impulse response (FIR) (FIR) filter, and it can compromise selection reach best between image quality and manufacturing cost.

At last, the single-processor of system also must be able to carry out foregoing level conversion.Fig. 6 shows the graphic representation of this process.It is as follows to produce output sample in conjunction with input sample:

Y1＝X1 Y2＝85/512X1+427/512X2 Y3＝171/512X2+341/512X3

Y4＝1/2X3+1/2X4 Y5＝341/512X4+171/512X5

Y6＝427/512X5+85/512X6 Y7＝X6

This system adopts 2608 instructions to carry out all above-mentioned functions, and it is about 95% of 2730 available instruction spaces.Advanced FIR filter failed to be applied to early stage SVP in the past, because they will waste instruction space amount.Usually, converting comprises and utilizes conversion factor as much as possible,, conversion factor has been done restriction here, any function that can sacrificial system.At last, this system has the performance that can be fit to new and different standards.In this system, the PALplus standard that was not considered in the past also can be used at an easy rate.This is because system has inherent adaptability.More than, although be that specific embodiment at a kind of improved PS based on the scan line processor is described, these descriptions are not to be used for limiting the scope of the invention, scope of the present invention is determined by appending claims.

Claims (9)

1. improved image processing system comprises:

Numeral input by a kind of common sample at rates;

A scan line image processor, it is characterized in that, described processor receives described numeral input by described common speed, and can operate motion detection, the preposition scanning interpolation of Motion Adaptive, the level and vertical conversion finished, and definition control is applied to described input.

2. the system as claimed in claim 1 is characterized in that, described common speed is 720 samplings of every row image input.

3. the system as claimed in claim 1 is characterized in that, described vertical conversion utilizes finite impulse response filter to carry out.

4. the system as claimed in claim 1 is characterized in that, described level converts and utilizes finite impulse response filter to carry out.

5. the system as claimed in claim 1 is characterized in that, described system can receive and show multiple radio and television standard.

6. the system as claimed in claim 1 is characterized in that, described numeral is input as NTSC.

7. the system as claimed in claim 1 is characterized in that, described numeral is input as the NTSC feeder connection.

8. the system as claimed in claim 1 is characterized in that, described numeral is input as PAL.

9. the system as claimed in claim 1 is characterized in that, described numeral is input as PLAplus.

Images