CN1276946A - Low noise encoding and decoding method - Google Patents
Low noise encoding and decoding method Download PDFInfo
- Publication number
- CN1276946A CN1276946A CN 97182481 CN97182481A CN1276946A CN 1276946 A CN1276946 A CN 1276946A CN 97182481 CN97182481 CN 97182481 CN 97182481 A CN97182481 A CN 97182481A CN 1276946 A CN1276946 A CN 1276946A
- Authority
- CN
- China
- Prior art keywords
- signal
- filtering
- resolution
- frame
- generates
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Landscapes
- Compression Or Coding Systems Of Tv Signals (AREA)
Abstract
An adaptive digital image processor precedes an MPEG2 encoder. The processor receives a high definition (1920 by 1080 pixels per image) video signal intended for broadcast or storage, and adaptively low-pass filters the signal. The signal is subjected to low-pass two-dimensional filtering to eliminate encoding artifacts and related noise. The video signal is then horizontally down-sampled to create a lower resolution hybrid signal (1280 by 1080 pixels per image). A receiver decodes and decompresses the hybrid signal. The hybrid signal is upsampled to its original resolution using existing hardware and software with a software modification.
Description
Invention field
The present invention relates to be used to reduce the Image Compression of digital video code bandwidth requirement.
Background of invention
FCC (FCC) ratified by Major Leagues (Grand Alliance, the GA) digital high-definition television of Ti Chuing (HDTV) standard, thereby clear the path for the ground digital television broadcast of the U.S..GA HDTV system has adopted Motion Picture Experts Group (MPEG 2) video compression standard.Details about this respect, please refer to " universal coding of information technology-moving image and related audio information: video ", (Information Technology-Generic Coding of Moving Picture and AssociatedAudio Information:Video, " ISO/IEC13818-2:1996 (E) .).Utilize modern and perfect video compression technology, for example, source processing, estimation and compensation, transmission represent and statistical coding, and MPEG compression system contracts and traffic bit speed can be reduced to original 1/50th or still less.A complete HD signal that reaches one second needed about 1 * 10 before compression
9The position.Specification requirement is recommended as GA, and 1920 * 1080 pixels (figure basic element) image that transmits with the speed of per second 60 fields is compressed into the data that the speed of per second 18 megabits transmits and uses for digital broadcasting.
The GA video compression system generally includes two main subsystems, and promptly preprocessor and MPEG 2 video encoders follow output buffer closely after them.Input to preprocessor is the analog video of rgb format.The preprocessor digital input signal, and each color component is carried out gamma (gamma) proofread and correct, with the nonlinear response of compensating image video camera.Gamma correction has reduced and has been included in compressed image, especially in the dark space of image, in the visibility of quantizing noise.Then, preprocessor transforms to SMPTE 240M YC1C2 color space with the RGB sample linearity of digitlization and gamma correction.At last, the chromatic component that finally obtains is carried out secondary sample and form 4: 2: 0 digital video input signal.Except just having described of task, preprocessor can also be implemented image transitions.For example, in digital broadcasting satellite system, vision signal is decimated into every line 544 pixels from every line 720 pixel level, further reduces bandwidth requirement.This signal is sent to MPEG 2 video encoders.
MPEG 2 video encoders compress input digital video signal by the time redundancy position between frame and some the spatial redundancy positions in the removal frame.In general, compression reaches by some kinds of technology recited above of continuous use.Adjusting quantified precision can make encoder generate with any compression position flow by the speed of using decision.Quantification in MPEG 2 systems is to realize at the DCT coefficient of data block, and it can be an original image information, or according to the remaining information of estimation.Utilize the quantization matrix that combines with the quantization step that can change in proportion, quantizer is only selected from each DCT piece and the DCT coefficient of quantification fraction supplies to send usefulness, causes the minimizing of valid data.Quantization matrix can be that unit changes with the frame according to the Distribution Statistics of DCT coefficient and the content of vision signal.For the zones of different in the frame, quantification can be the unit fine tuning with the macro block according to the complexity of macro block (macroblock), by changing quantization step in proportion.For given carry-out bit speed, output buffer provides the control signal of using for encoder, adjusts the quantization step that is used for particular frame so that maximize quantization resolution in the bandwidth that is suitable for.
In theory, video compression system has been removed radio-frequency component, when image by reconstruct with when showing because these radio-frequency components are lost, no longer viewed person feels.Remaining low-frequency component is quantized in the bandwidth range that is suitable for.Concerning the beholder, when image reconstruction, the quantizing noise that is brought in the signal also should be sightless.Yet, in the system of reality, between information that sends and quantization step, trade off about bandwidth.Enough coefficients about quantizing if system does not descend, system then increases the quantization step of the block artefact (blocky artifact) that causes in the reconstructed image.If image has been lost too many high-frequency information in compression process, reconstructed image then will have other tangible edge artefact (edge arfifact).
In addition, the quantitative differences between each frame makes the frame in the image sets (GOP) comprise variable radio-frequency component.For example, the I frame may contain a considerable amount of high frequency coefficients that descend in cataloged procedure, and P and B frame keep with the I frame in the corresponding radio-frequency component of radio-frequency component that descends.Now, because variation has taken place high-frequency information between the frame of mutual reconstruct being used for, the GOP of reconstruct has comprised artefact.
These problems appear in the current defined GA system.Compression HD picture signal has further reduced the quality of display image on the contrary.Owing to once have only a program to send by transponder, satellite broadcasting supplier is reluctant to send the HD signal.Up to now, in a satellite channel (for example, 24MHz 4-PSK modulation), fully compress the HD program at synchronization to be suitable for two picture qualities that program causes the beholder to be difficult to receive.Therefore, because channel uses satellite broadcasting supplier unwilling broadcasting HDTV inadequately.Equally, when a program took several SD programs fully and may reside in wherein channel, terrestrial broadcasting supplier also was unwilling to bear the responsibility that complete HD program is provided.
Summary of the invention
According to principle of the present invention, the type of Digital Image Processor identification video signal, and selectively original signal format is converted to different needed forms.After being filtered, the signal of conversion converts needed original form again to.That signal after the filtration is converted into low resolution and be compressed into the signal that transmits with targeted bit rates.At last, compressed signal is sent on the dateout channel.
The accompanying drawing summary
Fig. 1 shows a kind of structure according to video compressor of the present invention.
The detailed structure of the unit 22 in Fig. 2 displayed map 1.
Fig. 3 shows that a kind of of sef-adapting filter who is included in the unit 22 may respond.
Fig. 4 shows the flow chart that utilizes exemplary emission system of the present invention.
Fig. 5 shows the flow chart that utilizes exemplary receiving system of the present invention.
Preferred embodiment is described
Comprise two dimension (for example, the vertical and level) filter that the MPEG 2 that installs in accordance with the principles of the present invention is included in the encoder front.Each all produces the information that can be used for improving gross efficiency by other unit encoder, output buffer and filter.Such information relates to, and for example, the bit rate and the image texture structure of each unit selected, come to image motion, picture contrast, quantization matrix selection, scale factor.Information is to communicate by the controller that monitors encoding process or between by the unit that resides in the separate controller control in each unit.
The information that the controller evaluation arrives, and being identified in general character on image sets, a frame or the partial frame, this general character can advantageously be used for revising the operation of filter and/or encoder so that effectively image sets, frame or partial frame are encoded into and send with targeted bit rates.Because adjustment filter ratio adjustment encoder brings out noise still less, therefore, in general, all is to adjust filter.In addition, filter is actually one group of filter, adjusts each filter coefficient on demand, can reach maximum flexibility.These filters are horizontal antialiasing low pass filter, vertical low pass filters and two-dimensional low-pass filter, are being arranged in order according to the order that provided just now usually.Controller is assessed with respect to the reception information when pre-filter and encoder apparatus, and according to one or more outstanding general character one or more filters and/or encoder is made adjustment.Last result is, the filtered device of input signal according to make usually encoder whole with by the relevant image sets of the outstanding general character of the data of uniform encoding, a frame or partial frame on the mode low-pass filtering of uniform encoding image.
Code signal can send in the bandwidth that is suitable for, and is not having under the situation of the artefact that will occur under other condition by reconstruct and demonstration.For the high-definition signal with every picture frame 1920 * 1020 pixels, horizontal resolution is reduced to every line 1280 pixels before the coding after filtering, and has further compressed the bandwidth that sends signal.This result is a vision-mix resolution of utilizing less ground software alteration HD receiver just can receive, decode and show.
Demonstrative structure according to video compression system of the present invention is presented among Fig. 1.Incoming video signal is received by film detector 20, and whether it discerns this signal is film (film) signal that has been reformated into per second 30 frames by the telecine format conversion method from per second 24 frames.The movies signal of reformatting flows to the suitable part of adapting to image processor 22 as described below.If input signal is not the movies signal of reformatting, signal is sent to another part of adapting to image processor 22.The movies signal of identification telecine format conversion is finished by known method.
Processor 22 receives from output buffer 26 with from the control signal of MPEG 2 encoders 24 by controller 28, and the filtering image frame, makes encoder 24 can effectively this frame be encoded within the bit rate that is suitable for and does not have tangible artefact usually.Processor 22 is two dimension (2-D) (for example, level is with vertical) filtering signal on demand, improves the reconstructed image quality that is limited in MPEG 2 coding streams on the average bit rate.Its objective is according to the local 2-D frequency spectrum of with regard to image definition and coding artefact, the minimum mode of MPEG 2 reconstructed images infringement being revised image, improve MPEG 2 code efficiencies.The filtering of signal can be to predetermined data, and for example, image sets (GOP) or frame group, single frames or pixel of a pixel are carried out.
2-D filter low pass filtering image.Best is, the high-frequency information that remove concerning the beholder be redundant be again unessential.In fact, in order to obtain desirable bit rate, some are that important high-frequency information can be removed to the beholder.But, just as discussed below, comprised that before MPEG 2 codings the image of system's generation of processor 22 is better than the image that system generated that does not have processor 22.
Filtering signal is by MPEG 2 encoders 24 codings, and it receives the image parameter of from processor 22 and output buffer 26 by controller 28, and adjusts MPEG 2 compressions so that observe the bit rate that is suitable for.Compression is according to carrying out with the described same way as of GA specification requirement.Encoder 24 is sent to output buffer 26 with packed data.Buffer 26 provide utilize known signal treatment technology coding, modulation and send by transmitting channel, with the packed data to be transmitted of set rate.The modulation before, compressed signal can be sent to statmux, carry out with multiplication routine multiplexed so that on individual channel, send.The signal processing unit of buffer 26 back is well-known, therefore, for graphic simplicity, is not presented among Fig. 1.
Video compression system can be constructed to be permeable to receive the vision signal of any kind.The system of Fig. 1 is configured to receive TV (projection TV) and film (film) program that is formatted into known industry standards.For the system of Fig. 1, a kind of general structure can be for example, to receive the output from the preprocessor described in the background technology of front.By adding suitable hardware and/or software, system also can be constructed to be permeable to receive the vision signal of other type.These structures all do not show so that reduced graph 1.
Can be used for improving the existence of certain relation of code efficiency in the film detector 20 identification input signals: 30 frames of lining by line scan (Class1) 60 fields/second interlacing scan source, (type 2) 60 field/seconds interleaved 30 frames/second film, (type 3) 60 field/seconds interleaved 24 frames/second film, (type 4) progressive-source, (type 5) 60 frame/seconds/second film and 24 frames of lining by line scan (type 6) 60 frame/seconds/second film.Detection is according to the external control signal (not shown) or by carrying out as the employed known technology of current standard definition (SD) MPEG 2 encoders.As described below, signal format information is sent to adapting to image processor 22 with signal.Film detector 20 also detect this signal be the interlacing scan type or line by line scan type, and this information is sent to processor 22.These scan types are exemplary, and determine to make the parameter of signal flow to processor 22.Embodiment about other field and frame rate also can be used.
Adapting to image processor 22 realizes that several reductions will be by the programmable functions of the quantity of the data of encoder 24 compressions.Processor 22 is operated each frame usually, causes treated frame can obtain forced coding, and eliminating or reducing the beholder widely is obvious noise.Usually processor 22 can be regarded as spatially-variable 2-D low pass filter, (down-sample) each picture frame and filtering is selected from signal adaptively 2-D radio-frequency component because it is taken a sample up and down from the space.Adaptive-filtering can be adjusted to and carry out on the row frame, carrying out in single frame or pixel ground of pixel carries out, and generates treated frame.
Processor 22 can convenient coding to any kind signal.But, embodiment hereto, processor 22 is programmed to be operated the HD data of GA specification requirement defined.This both can be that every width of cloth image 1920 * 1080 pixels also can be every width of cloth image 1280 * 720 pixels.According to this GA specification requirement, each HD call format is broadcast about per second 18 megabits.To oversimplify in order making to discuss, only 1920 * 1080 forms to be gone through now.This discussion is equally applicable to 1280 * 720 forms or any other form.
Fig. 2 has shown the detailed structure of adapting to image processor 22.According to the signal format information that receives from detector 20, picture signal is routed to interlacing-progressive converter 221 (Class1), contrary telecine format conversion (de-telecine) unit 222 (type 2,3) or is not added by controller 28 (Fig. 1) and is sent to space limit low-pass filter 223 with revising.After unit 221 and 222 had carried out signal handling, filter 223 received their output.
If signal format comprises 60Hz interlacing field, transducer 221 just receives this signal, and converts it the progressive frame of per second 60 frame rate to.Progressive frame comprises all images information in every frame.The artefact that the filtering progressive-scan signal may occur when not bringing the field information of filtering interlace signal usually into.Transducer 221 utilizes known method to convert the interlacing field to progressive frame.
Contrary telecine format conversion unit 222 is eliminated the redundant field and the original film of lining by line scan of reconstruct of 60Hz interlacing scan film.Progressive format makes the do not moved interference of artefact of subsequently vertical low pass.Just in case film source (type 2 or type 3) input is taken as the source input of Class1 and handles, vertical low pass will reduce the ability that MPEG 2 encoders detected and suitably handled the film source content.Code efficiency is also with sustain damage.Unit 222 becomes the single-frame signal of formula with conversion of signals, and removes redundant field/frame before filtering, because filtering may filter away redundant information with having nothing in common with each other.If redundant information was not removed before filtering, information may be different after filtering so, encoder may not be identified as this signal the signal of type 2/3, and then, encoder will be encoded in other cases because the removed information of its redundancy.
In addition, by the design that single output clock can simplified processor 22 is provided on unit 222.If unit 222 provides per second 24 frame rate and two kinds of outputs of cine-oriented image line by line of per second 30 frame rate, then need two output clocks and support circuit.
Signal according to the original generation of per second 30 frame progressive formats directly is sent to filter 223.Filter 223 requires to be expressed as the video information of complete graph picture frame.Spatial low-pass filter 223 is actually one group of filter.For example, as mentioned above, first filter is the horizontal low pass filter of antialiasing, and second filter is that vertical low pass filters and last filter are the 2-D low pass filters.The tap coefficient of each filter can be according to being provided with adaptively from as shown in Figure 1 the encoder 24 and the control information of buffer 26.Progressive signal is by horizontal low-pass filtering, so that from obscuring to eliminate the transducer 226 of taking a sample under the sampling rate subsequently.As discussed below, the output of the final level of every line 1920 pixels will become every line 1280 pixels.In order to eliminate the noise of obscuring in the last signal, low pass filter 223 has the cut-off frequency in every 640 weeks of line.The horizontal antialiasing filter that is included in the unit 223 can be 17 tap finite impulse response (FIR) (FIR) filters that have following tap coefficient:
[f
0、f
1、...、f
15、f
16]=[-4、10、0、-30、48、0、-128、276、680、276、
-128、0、48、-30、0、10、-4]/1024
The coding HD vision signal that bit rate has reduced needs additional vertical low pass usually, with the bandwidth of further compressed video signal.Removing the vertical high frequency energy before mpeg encoded is necessary to obtaining receivable overview image quality.The most vertical frequency district of high phse sensitivity is attenuated.Vertical cut frequencies is configured to certain branch several times of Nyquist (Nyquist) frequency.For example, for some video contents (material), half the cut-off frequency that approximates the wire rate of HD input signal greatly may be suitable.For the HD signal that has 1080 lines/picture altitude (l/ph), this will be corresponding to the cut-off frequency of 540l/ph.This frequency can be programmable, and programmable cut-off frequency will be by controller 28 according to determining from the encoder 24 of Fig. 1 and the suitable parameter of buffer 26 (that is, desirable bit rate, quantization matrix, etc.).The vertical low pass filters that is included in the unit 223 can be the 17 tap FIR filters that have following tap coefficient:
[f
0、f
1、...、f
15、f
16]=[-4、-7、14、28、-27、-81、37、316、472、316、
37 ,-81 ,-27,28,14 ,-7 ,-4]/1024 as an optional value, cut-off frequency can equal the twice of the wire rate of SD signal.In general, vertical low pass filters is right after the back at horizontal antialiasing filter.
Processor 22 carries out vertical filtering, rather than vertically selects, thereby has kept constant vertical line resolution.Usually for interlaced video signal, filtering is better than selecting.The vertical line resolution of changing relevant interlaced image sequence needs complicated hardware and software, causes the receiver cost to raise.Because improved and Nyquist (Nyquist) sampling (that is, not having sampling) relevant tap complexity, the vertical sampling rate transition will be taken into account the vertical high frequency performance.The cost consideration of receiver is not encouraged to reduce vertical resolution usually reduce artefact and encoded bit rate.In the vertical sampling rate converter that replaces aforesaid vertical low pass filters, use prior art and can make the remarkable variation of image quality.But, effectively and have the vertical sampling rate converter of cost advantage can replace vertical filter as herein described, and do not depart from principle of the present invention.
If necessary, can revise by above software, and be applied on the pixel level to obtain the targeted bit rates that in reconstructed image, does not generate artefact about the coefficient of level and vertical low pass filters.In general, be that the coefficient modifying of unit is just enough with the frame.A kind of possibility that is used for slow processor is to be several different coefficient sets of filter pre-programmed and the image information of selecting only that group to be used for handling.With respect to the system that does not have sef-adapting filter, the big flexibility of sef-adapting filter makes whole system can produce the data flow that has less artefact.
At signal by unit 223 after horizontal direction and vertical direction are carried out low-pass filtering, controller 28 judges that whether this signal is unit uniform encoding and do not bring tangible quantizing noise into the frame by encoder 24.If so, signal is sent to unit 224,225 or 226 according to its form, will discuss to this below.But if cataloged procedure might be brought into noise and/or artefact in the signal, the two-dimensional low-pass filter that then passes the signal in the unit 223 carries out further adaptive-filtering.The Control Parameter of from processor 22, encoder 24 and output buffer 26 (Fig. 1) makes controller 28 (or the independent cell controller in the unit 223) need judge whether further filtering.The Control Parameter that is used to make this judgement is, for example, and the motion and the measurement result of contrast, suitable quantization table, code efficiency and current goal bit rate.
2-D filter in the unit 223 has reduced from mainly along diagonal, and is not only the high-frequency information along the picture frame of level or vertical direction.With respect to diagonal, human eye is especially responsive to the high-frequency noise along vertical and horizontal direction.Remove enough high-frequency informations along diagonal and generate the higher quality that has less discernable noise usually by encoder 24 uniform quantizations.The same with the filtering before all, diagonal filter is operated the entire image frame, and is programmable.
Diagonal filter can with the quantization matrix compatibility in the encoder.The quantization matrix that is used for the quantification of I frame often uses the diamond shaped matrix.But, because B and P frame use the quantization matrix of other type, kept radio-frequency component in compression that they carry out and the motion compensation process process in encoder 24, therefore, these diamond shaped matrixes often bring out noise.MPEG 2 encoders 24 processed data into I, P and B frame in the estimation network before, the filter of processor 22 was removed high-frequency information from each picture frame.Therefore, radio-frequency component is not only eliminated from the I frame but also from P and B frame usually.When reconstruct, image does not contain the known artefact that is produced by MPEG 2 codings usually.
With reference to Fig. 1, in fact, controller 28 is estimated signal parameter (for example, motion, contrast etc.) before the given frame of processor 22 filtering, and determines the system's setting about all filters in the unit 223, comprises needed diagonal filtering.In the Filtering Processing process to given frame, controller 28 monitors the signal parameter of from processor 22, encoder 24 and buffer 26, and changes coefficient on demand to keep making the targeted bit rates of artefact/noise minimum.Each frame is to carry out filtering according to nearest signal parameter, and is sent to encoder 24 and compresses, and is sent to buffer 26 then and is input in the filter 223 as follow-up.
If signal comes from the movies signal of per second 24 frames, then filtering signal is applied on 3: 2 drop-down unit (pull down unif) 224.Unit 224 duplicates selected frame so that the output signal of per second 30 frames to be provided.This utilizes known method to finish.Then, signal 224 is sent to sampling switch 226 under the level from the unit.
Field secondary sample unit 225 will become the interlacing scan form from progressive scan format conversion from the progressive signal of filter 223.This conversion is finished by known method.Because the progressive frame speed from unit 221 is 60Hz, need not to be converted back to interlaced format, this signal will contain the data volume of twice.Interlace signal is applied to transducer 226.
Sampling rate conversion device 226 directly receives the progressive signal of per second 30 frames from filter 223.In addition, as mentioned above, unit 224 and 225 also offers signal transducer 226.Transducer 226 will be taken a sample into selected transmission form under the HD signal.This form also needs not be a kind of reference format.It can be any image ratio and desirable frame size.But noncanonical format needs receiver that it is made amendment.
When transducer 226 received 1920 * 1080 GA HDTV signals, 226 times sampling level information of transducer were also exported 1280 * 1080 mixed pixel frame format.The compatible receiver of GA HDTV can receive the picture frame that contains 1920 * 1080 pixels and 1280 * 720 pixels.Therefore, can be to the compatible receiver correct of GA, to support 1280 pixel level resolution and 1080 pixel vertical resolutions.The raising of compatible receiver combination of hardware vertical resolution will be taken a sample into 1920 horizontal pixels on 1280 horizontal pixels.But, can not require GA subordinate receiver GA subordinate receiver can not be programmed to and can receive the picture frame resolution of 1280 * 1080 pixels of form (level * vertical) as defined.Hardware is on the position that receives the decode this resolution, but must add decoding and the support software of the resolution of only improving the standard.With respect to redesign with add the needed new hardware of other noncanonical format, add software and be not only simply but also cheap.
Because existing Display Technique can't show the resolution of 1920 pixels/line, processor 22 provides 1280 * 1080 mixed format.At present, best televimonitor also can only show the resolution of about 1200-1300 pixels/line.Therefore, the output resolution ratio of horizontal direction is limited in 1280 pixels/line what negative effect is picture quality almost do not had.By the display resolution (1280 * 1080) that is subjected to existing decoding and decompression receiver hardware supports is provided, owing to only software is done necessary change, receiver manufacturer will be subjected to minimal effects.For some receiver, for example be used for the receiver of broadcasting satellite, software modification is link remote download and installation via satellite.For these receivers, need not the service technology expert and participate in.
Because ground and satellite programming provide the unwilling transmission in commercial city HD program, this mixed format has superiority.Satellite repeater sends the data flow of about per second 24 megabits (Mbps).HDTV broadcasting in ground can send the data flow up to 19Mbps, comprises HD program and the out of Memory (for example, audio frequency, TV guide, condition rating etc.) of 18Mbps.Each can only transmit a HDTV program at most present satellite repeater, this just satellite programming provider complain nonprofit focus place.Simply horizontal frame resolution being reduced to 1280 from 1920 is not sufficient to make single satellite repeater to send two HD programs simultaneously.The filtering that is provided by processor 22 helps carrying out so two HD transmissions on individual channel.
The filtering characteristic that is provided by processor 22 can have different shape, comprise diamond shape, cross shaped head and hyperbolic shape from one to another that, for each filter, filtering is the diagonal angle here.A possible shape, the 2-D hyperbola, should with in especially possess some good points, it has amplitude-frequency response as shown in Figure 3.Adjustable filtering cut-off frequency is configured to make selected GOP, frame or partial frame evenly to be compressed by encoder 24 usually.If necessary, can carry out filtering to level and vertical high frequency information in addition, but this is unnecessary usually.Along with the change of image complexity, or along with the raising of applicable bit rate, can reduce by the data volume of the filter filtering of diagonal filter and other front.The 2-D filter can be called, and for example, all has the two-dimentional FIR filter (13 * 13) of 13 taps or 2-D finite impulse response (FIR) (IIR) filter in each direction.
The 2-D FIR filter that is included in the unit 223 can be 13 * 13 tap filters that have following tap coefficient:
0 0 0 1 -3 5 -5 5 -3 1 0 0 0
0 0 1 -1 -3 9 -11 9 -3 -1 1 0 0
0 1 -1 -4 4 6 -13 6 4 -4 -1 1 0
1 -1 -4 1 14 -13 4 -13 14 1 -4 -1 1 -3 -3 4 14 5 -47 61 -47 5 14 4 -3 -3
5 9 6 -13 -47 -55 193 -55 -47 -13 6 9 5 -5 -11 -13 4 61 193 556 193 61 4 -13 -11 -5
5 9 6 -13 -47 -55 193 -55 -47 -13 6 9 5 -3 -3 4 14 5 -47 61 -47 5 14 4 -3 -3
1 -1 -4 1 14 -13 4 -13 14 1 -4 -1 1
0 1 -1 -4 4 6 -13 6 4 -4 -1 1 0
0 0 1 -1 -3 9 -11 9 -3 -1 1 0 0
0 0 0 1 -3 5 -5 5 -3 1 0 0 0
For these coefficients, the DC gain is 1024.These coefficients present the octant symmetry, and it has provided 28 independent coefficients.The symmetry coefficient district comparatively fast is provided with tunable filter.But, for example,, then might each octant have nothing in common with each other if filtered image or zone present different characteristics in certain part of image.
The filter response of processor 22 can be that unit varies continuously to another group coefficient from one group of coefficient with the pixel to pixel.Therefore, as discussed below, under the rate constraint on the throne, processor can present different running parameters to keep high quality image.
As previously mentioned, processor 22 can obtain adaptively revising according to the parameter that is used for defining the filter adaptivity and carry out adaptive-filtering.For example, the variance in the picture frame can be used for image is divided into several zones to carry out different processing.The edge is the important images feature, and that regional encoding error and they also can define image range because main edge is in covert.Colourity can be used for discerning the zone of low-complexity, resembles skin and sky etc.Also can discern and handle the texture in a zone.Texture does not have the edge so important usually.Therefore, texture is used for discerning than other regional filtering and gets more zone.In addition, movie contents also can be used for important portrait or the action of positioning requirements than high coding efficiency, therefore less filtering.Background is softening by the field depth of camera lens usually, can be filtered away more.(pan) taken in servo-actuated can be used for determining that with scanning information the center of image of interest supplies processor 22 to carry out different processing usefulness.
The work of encoder 24 and MPEG 2 operating suchs.Encoder 24 can provide processor 22 can be used for improving the information of performance by controller 28.Such information can comprise, for example, and bit rate information.This bit rate information can comprise average bit rate, framing bit speed and macro block or the piece bit rate of GOP.Other information that can improve processor 22 performances comprises the type of the complexity of discrete cosine transform, the quantization matrix that using and the quantization matrix step-length of using.In addition, processor 22 also can offer encoder 24 with information by controller 28, and the operating state of adjusting encoder is to improve coding efficiency.
After utilizing known technology formation transmitting grouped data stream, the HD signal for example, as the described mode of Grand Alliance specification requirement, sends to receiver according to known manner.Except on the receiver complete HD pixel resolution being required the sampling, the signal processing that is provided by processor 22 is transparent to the decoder in the compatible receiver of Grand Alliance.
On receiver, utilize known technology that data stream is separated mediation transport stream is handled recovery data packets and programme information.For the HD program according to mixed format recited above, if display requires complete HD signal, then signal is taken a sample on the horizontal direction quilt in video-stream processor.Vertical line number in the picture signal does not change.The complete HD signal that so just reconstructs 1920 * 1080 resolution shows for the high-definition image reproducing apparatus to be used.If image display requires to be lower than the signal of complete HD signal, this signal utilizes known method to carry out sampling down suitably in the image reconstruction procedure before showing.The compatible receiver of existing Grand Alliance requires software is made amendment, so that can the reconstruct mixed signal.Software modification makes the level that is assigned to Grand Alliance standard mode can obtain independently according to the needs of input signal selecting with vertical hardware and software handling procedure.
Fig. 4 is the flow chart that flows through the video signal of coded system.In step 30, signal format is discerned and identifying information is transmitted with vision signal.Format information can represent for example, whether signal is that film and the signal that is derived from per second 24 frame formats is interleaved or lines by line scan.If vision signal is an interleaved signal,, conversion of signals is become the progressive-scan signal of per second 60 frames then in step 31.If signal by 3: 2 drop-down cell translation, then in step 32, is removed redundant frame.If vision signal has been the projection TV progressive-scan format, then directly be sent to the filter of step 34.In step 34, vision signal is by low pass spatial filtering.Low pass spatial filtering comprises aforesaid vertical, level and diagonal filtering.Step 35 will be converted back to interleaved signal again at the progressive-scan signal that step 31 is changed from interleaved signal.In step 36, signal is carried out drop-down processing in 3: 2 to restore the redundant frame of step 32 deletion in front.The vision signal that directly is sent to step 34 from step 30 directly is sent to step 38 from step 34 now above.In step 38, become to mix 1280 * 1080 HD signal resolution as defined above by secondary sample with one of 36 vision signal from step 34,35, or another kind of desirable output format.In step 40, by foregoing mode MPEG 2 coding mixed signals.Step 42 transmission process code signal and step 44 are according to requiring modulated transmission signal by the transmission such as the such delivery channel of RF terrestrial broadcasting channel.At last, step 46 sends modulation signal.Step 42-46 realizes by known method.
Fig. 5 is the flow chart that flows through the emission picture signal of receiver.The image resolution ratio of this flow chart supposition received signal is mixing 1280 * 1080 HD signal resolution as defined above.In step 50, send signal and receive and obtain demodulation by tuner.In step 52, restituted signal is carried out MPEG 2 decodings and decompression.Step 54 is differentiated vision signal by the control information that transmits with signal and is identified as each image 1280 * 1080 pixel.Grand Alliance MPEG 2 agreements will send signal and offer image resolution ratio information.The same with the resolution of any other definition, mixed signal is discerned by private code usually.In other cases, mixed signal also can be passed through, and for example, the information in the user data of transmission data is determined.In step 56, mixed video signal is taken a sample into complete 1920 * 1080 HD signals on by level in showing processing procedure.Mixed signal is to utilize new software relevant with existing hardware and the software that resides in the foregoing receiver to go up sampling.At last, in step 58, the HD vision signal is presented on 1920 * 1080 displays fully.Step 50,52 and 58 has been used known method.
The image reconstruction that apparatus and method recited above can be applied to be improved in many structures shows for high definition to be used.Can use the possibility of adaptive and non-self-adapting according to the requirement decision of concrete system.Several in these possibilities are discussed below.
The non-self-adapting strategy is that the frame filtering with processor 22 is limited on the targeted bit rates and makes all images obtain uniform treatment.Another kind of non-self-adapting strategy is that the center of hypothesis display image is most interested zone.This also supposed concerning the beholder image around very not interesting, therefore not too important yet.The filter factor of processor 22 is all to obtain uniform treatment by controller 28 by the image informations that be provided with and all of the function parameters as the pixel space position.
The self-adaptation selectable scheme is to utilize based on the texture model parameters of source images, local video variance, colourity or other image complexity degree image to be divided into several zones.The filtering characteristic of processor 22 is to revise at different region adaptivity ground.
Another kind method is the filtering characteristic of revising adaptively as the function of the difference of actual bit speed and targeted bit rates.In this case, single parameter control is about the conversion of the filter factor of 2-D frequency response.
Another kind of strategy is that the 2-D frequency response of filtering that processor 22 is provided is designed to and encoder 24 employed quantization matrix compatibilities.Quantization matrix can be seen as the low pass filter with 2-D shape.For this strategy, the value of filter factor will be the function of quantization matrix step-length.Along with step-length changes according to known encoder operation, corresponding change also can take place in corresponding filter factor.
Possibility recited above has illustrated the flexibility of using the system of the principle of the invention.Such system is preferably in the interior operation of scope of MPEG 2 rate controlled, so that expand the compressed capability of MPEG 2 by reducing coding artefact and other noise.Versatility and economic and practical that HDTV uses are that the application of the invention is improved.The HD number of programs that each transponder sends in direct broadcast satellite system (that is, 24 MHz 4-PSK) is increased to two programs from a program, or has a HD program of a plurality of SD programs.The ability principle according to the present invention that sends a HD program that has a plurality of SD programs in 6 MHz terrestrial broadcasting channels reaches.In the past, the broadcasting station was limited in and only sent a HD program in the channel, or only sent a plurality of SD programs in a channel.
Although the HD signal send and receiving system in describe the present invention, principle of the present invention can be applied to during such as data-storage system other install.In such as the such system of data video disc (DVD), coding and stored video data are used for resetting later on.This storage medium has only limited applicable memory space.If coded program, film or other video sequence surpass the free space on the medium, will further encode/compress is complementary program and free space, may cause like this and make us the artefact that is difficult to receive.Invention recited above can be used for effectively program being encoded into have make program with the dish be complementary than low bitrate.Perhaps, now, a plurality of programs can be installed on the dish.The stored digital that stores on the tape also can have above-mentioned advantage.
Claims (22)
1. a processing presents the method for first and second vision signals of the first and second different images forms respectively, comprises detecting the step that described first or second vision signal exists and further comprising the steps:
A. when described first vision signal is detected,
(a) convert described first vision signal to different form, generate switching signal;
(b) the described switching signal of filtering generates filtering signal;
(c) described filtering signal is converted back to again the original form of described first signal, generates switching signal again;
(d) described switching signal again is converted to low resolution, generates the low resolution signal;
(e) the described low resolution signal of coding generates code signal; With
(f) described code signal is sent to delivery channel; And
B. when described second vision signal is detected,
(g) described second vision signal of filtering generates filtering signal;
(h) described filtering signal is converted to low resolution, generates the low resolution signal;
(i) the described low resolution signal of coding generates code signal; With
(j) described code signal is sent to delivery channel.
2. the method for claim 1, wherein:
Described first vision signal is that interleaved signal and described interleaved signal are converted into progressive-scan signal in step (a).
3. the method for claim 1, wherein:
Described first vision signal is to be converted into signal through contrary telecine format conversion through the movies signal of telecine format conversion and described movies signal through the telecine format conversion in step (a).
4. the method for claim 1, wherein:
Described second vision signal is a progressive-scan signal.
5. the method for claim 1, wherein:
Described filter step provides low-pass filtering.
6. method as claimed in claim 5, wherein:
Described filter step provides two-dimensional filtering.
7. the method for claim 1, wherein:
Described filter step provides the adaptive-filtering that is applicable to one of a picture group picture frame, single frames and partial frame.
8. the method for claim 1, wherein:
Described filter step provides the characteristic adaptively modifying filtering characteristic of time low-pass filtering and the described signal of response.
9. the method for claim 1, wherein:
Described filter step provides the characteristic adaptively modifying filtering characteristic of low pass spatial filtering and the described signal of response.
10. the method for claim 1, wherein:
Described coding step is MPEG 2 compatibilities.
11. the method for claim 1, wherein:
Described low resolution signal has the resolution of every frame 1280 * 1020 data samples.
12. the method for claim 1, wherein:
Described first and second vision signals are the high-definition signals with every frame 1920 * 1080 data samples.
13. handle interlaced video signal and, comprise the following steps: for one kind through the method for one of film format signal of telecine format conversion
Detect the existence of one of described interlaced video signal and described film signal through the telecine format conversion,
Respectively described detection signal is converted to one of signal of progressive-scan signal and the contrary telecine format conversion of warp, generate switching signal;
The described switching signal of filtering generates filtering signal;
Respectively described filtering signal is converted back to again one of interleaved signal and telecine format conversion signal, generates switching signal again;
Described switching signal again is converted to low resolution, generates the low resolution signal;
The described low resolution signal of encoding generates code signal; With
Described code signal is sent to delivery channel.
14. method as claimed in claim 13, wherein:
Described filter step is that low-pass filtering and described coding step are MPEG 2 codings.
15. method as claimed in claim 13, wherein:
Described low resolution signal has the resolution of every frame 1280 * 1020 data samples.
16. a method of handling the progressive-scan video signal of no telecine format conversion comprises the following steps:
The described detection signal of filtering generates filtering signal;
Described filtering signal is converted to low resolution, generates the low resolution signal;
The described low resolution signal of encoding generates code signal; With
Described code signal is sent to delivery channel.
17. method as claimed in claim 16, wherein:
Described filter step is that low-pass filtering and described coding step are MPEG 2 codings.
18. method as claimed in claim 16, wherein:
Described low resolution signal has the resolution of every frame 1280 * 1020 data samples.
19. in the high definition video signal treatment system, a kind of present the resolution that comprises every frame 1280 * 1020 data samples more than a kind of image resolution ratio condition under handle the method for the digital video signal that receives, comprise the following steps:
The described signal generating solution coded signal of decoding;
Determine the image resolution ratio of described decoded signal;
If described decoded signal has the horizontal image resolution of every line 1280 samples, will convert different resolution from the horizontal information of described decoded signal to, generate switching signal; With
Described switching signal is sent to output equipment.
20. method as claimed in claim 19, wherein:
Described conversion is that upward conversion and described different resolution are every line 1980 horizontal samples.
21. method as claimed in claim 19, wherein:
Described conversion is that conversion down and described different resolution are low resolutions.
22. method as claimed in claim 19, wherein:
Described receiving digital video signal is MPEG 2 compatibilities.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN 97182481 CN1133327C (en) | 1997-12-23 | 1997-12-23 | Low noise encoding and decoding method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN 97182481 CN1133327C (en) | 1997-12-23 | 1997-12-23 | Low noise encoding and decoding method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN1276946A true CN1276946A (en) | 2000-12-13 |
CN1133327C CN1133327C (en) | 2003-12-31 |
Family
ID=5178528
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN 97182481 Expired - Lifetime CN1133327C (en) | 1997-12-23 | 1997-12-23 | Low noise encoding and decoding method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN1133327C (en) |
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1319372C (en) * | 2002-08-14 | 2007-05-30 | 瑞昱半导体股份有限公司 | Method and device for sampling digital image |
CN100388792C (en) * | 2006-04-24 | 2008-05-14 | 李博航 | Image processing method of increasing information encoding-decoding efficiency |
CN101765012B (en) * | 2010-01-20 | 2012-05-23 | 李博航 | Image information processing method |
CN101771877B (en) * | 2010-01-20 | 2012-07-25 | 李博航 | Information processing method |
CN101742320B (en) * | 2010-01-20 | 2012-09-12 | 李博航 | Image processing method |
CN103098468A (en) * | 2010-09-10 | 2013-05-08 | 汤姆逊许可公司 | Video encoding using block-based mixed-resolution data pruning |
US9338477B2 (en) | 2010-09-10 | 2016-05-10 | Thomson Licensing | Recovering a pruned version of a picture in a video sequence for example-based data pruning using intra-frame patch similarity |
US9544598B2 (en) | 2010-09-10 | 2017-01-10 | Thomson Licensing | Methods and apparatus for pruning decision optimization in example-based data pruning compression |
US9813707B2 (en) | 2010-01-22 | 2017-11-07 | Thomson Licensing Dtv | Data pruning for video compression using example-based super-resolution |
-
1997
- 1997-12-23 CN CN 97182481 patent/CN1133327C/en not_active Expired - Lifetime
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1319372C (en) * | 2002-08-14 | 2007-05-30 | 瑞昱半导体股份有限公司 | Method and device for sampling digital image |
CN100388792C (en) * | 2006-04-24 | 2008-05-14 | 李博航 | Image processing method of increasing information encoding-decoding efficiency |
CN101765012B (en) * | 2010-01-20 | 2012-05-23 | 李博航 | Image information processing method |
CN101771877B (en) * | 2010-01-20 | 2012-07-25 | 李博航 | Information processing method |
CN101742320B (en) * | 2010-01-20 | 2012-09-12 | 李博航 | Image processing method |
US9813707B2 (en) | 2010-01-22 | 2017-11-07 | Thomson Licensing Dtv | Data pruning for video compression using example-based super-resolution |
CN103098468A (en) * | 2010-09-10 | 2013-05-08 | 汤姆逊许可公司 | Video encoding using block-based mixed-resolution data pruning |
US9338477B2 (en) | 2010-09-10 | 2016-05-10 | Thomson Licensing | Recovering a pruned version of a picture in a video sequence for example-based data pruning using intra-frame patch similarity |
US9544598B2 (en) | 2010-09-10 | 2017-01-10 | Thomson Licensing | Methods and apparatus for pruning decision optimization in example-based data pruning compression |
CN103098468B (en) * | 2010-09-10 | 2017-02-15 | 汤姆逊许可公司 | Video encoding using block-based mixed-resolution data pruning |
Also Published As
Publication number | Publication date |
---|---|
CN1133327C (en) | 2003-12-31 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US7800690B2 (en) | Low noise encoding and decoding apparatus and method | |
RU2201654C2 (en) | Low-noise coding and decoding method | |
US20200374565A1 (en) | Image processing device and image processing method | |
CN1144468C (en) | Method for partioning and encoding data | |
CN1140126C (en) | System for encoding and decoding layered compressed video data | |
KR100481572B1 (en) | Time and Resolution Layering in ATVs | |
US8463056B2 (en) | Information processing apparatus and information processing method that decode a portion of a code stream | |
EP0712251A2 (en) | Method and apparatus for partially recompressing digital signals | |
CA2150467C (en) | Picture-in-picture tv with insertion of a mean only frame into a full size frame | |
RU2000119786A (en) | METHOD FOR LOW NOISE CODING AND DECODING | |
US6345388B1 (en) | Method and apparatus for transmitting and receiving television signal | |
KR100591211B1 (en) | Low noise encoding and decoding method | |
CN1133327C (en) | Low noise encoding and decoding method | |
CN1906946A (en) | Device and method for playing back scalable video streams | |
EP1239679A2 (en) | Low noise encoding and decoding apparatus and method | |
AU2003248178B2 (en) | Low noise encoding and decoding apparatus and method | |
KR100647202B1 (en) | Low noise encoding and decoding method | |
JP4724240B2 (en) | How to process a video signal | |
MXPA00006182A (en) | Low noise encoding and decoding method | |
JPH10112857A (en) | Method and device for coding and decoding image signal | |
PL185157B1 (en) | Apparatus for and method of low-noise encoding and decoding | |
Ng | A digital augmentation approach to HDTV | |
Reimers et al. | JPEG and MPEG Source Coding of Video Signals |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C14 | Grant of patent or utility model | ||
GR01 | Patent grant | ||
CX01 | Expiry of patent term | ||
CX01 | Expiry of patent term |
Granted publication date: 20031231 |