Classifications

• • H—ELECTRICITY
  • H03—ELECTRONIC CIRCUITRY
  • H03H—IMPEDANCE NETWORKS, e.g. RESONANT CIRCUITS; RESONATORS
  • H03H17/00—Networks using digital techniques
  • H03H17/02—Frequency selective networks
  • H03H17/0248—Filters characterised by a particular frequency response or filtering method
  • H03H17/028—Polynomial filters
• • H—ELECTRICITY
  • H03—ELECTRONIC CIRCUITRY
  • H03H—IMPEDANCE NETWORKS, e.g. RESONANT CIRCUITS; RESONATORS
  • H03H17/00—Networks using digital techniques
  • H03H17/02—Frequency selective networks
  • H03H17/0219—Compensation of undesirable effects, e.g. quantisation noise, overflow
• • H—ELECTRICITY
  • H03—ELECTRONIC CIRCUITRY
  • H03H—IMPEDANCE NETWORKS, e.g. RESONANT CIRCUITS; RESONATORS
  • H03H17/00—Networks using digital techniques
  • H03H17/02—Frequency selective networks
  • H03H17/06—Non-recursive filters
  • H03H17/0621—Non-recursive filters with input-sampling frequency and output-delivery frequency which differ, e.g. extrapolation; Anti-aliasing
• • H—ELECTRICITY
  • H03—ELECTRONIC CIRCUITRY
  • H03H—IMPEDANCE NETWORKS, e.g. RESONANT CIRCUITS; RESONATORS
  • H03H17/00—Networks using digital techniques
  • H03H17/02—Frequency selective networks
  • H03H17/06—Non-recursive filters
  • H03H17/0621—Non-recursive filters with input-sampling frequency and output-delivery frequency which differ, e.g. extrapolation; Anti-aliasing
  • H03H17/0635—Non-recursive filters with input-sampling frequency and output-delivery frequency which differ, e.g. extrapolation; Anti-aliasing characterized by the ratio between the input-sampling and output-delivery frequencies
  • H03H17/065—Non-recursive filters with input-sampling frequency and output-delivery frequency which differ, e.g. extrapolation; Anti-aliasing characterized by the ratio between the input-sampling and output-delivery frequencies the ratio being integer
  • H03H17/0657—Non-recursive filters with input-sampling frequency and output-delivery frequency which differ, e.g. extrapolation; Anti-aliasing characterized by the ratio between the input-sampling and output-delivery frequencies the ratio being integer where the output-delivery frequency is higher than the input sampling frequency, i.e. interpolation
• • H—ELECTRICITY
  • H03—ELECTRONIC CIRCUITRY
  • H03H—IMPEDANCE NETWORKS, e.g. RESONANT CIRCUITS; RESONATORS
  • H03H17/00—Networks using digital techniques
  • H03H17/02—Frequency selective networks
  • H03H2017/0298—DSP implementation

Definitions

• the present invention relates to an interpolation method and an interpolation apparatus for a digital signal, and in particular, to an interpolation method and an interpolation apparatus that can effectively reduce quantization noise even in a digital signal having large quantization noise like a case of decompressing compressed data.
• interpolation processing is performed by not only expanding the quantization bit length of the digital signal, read from the recording medium, by the predetermined bit length toward a low-order side, but also inputting the digital signal to a variable low pass filter whose frequency characteristic (cutoff frequency) dynamically changes according to an input waveform as shown in FIG. 22.
• a variable low pass filter whose frequency characteristic (cutoff frequency) dynamically changes according to an input waveform as shown in FIG. 22.
• a comparatively expensive recording medium such as semiconductor memory
• it is made to store a digital signal after compressing the digital signal so as to further effectively use memory capacity.
• an irreversible compression method that has a high compressibility but cannot completely restore a waveform to its original waveform
• the original signal is a sinusoidal wave as shown in FIG. 23( 1 )
• an original digital signal after A/ D conversion becomes as shown in FIG. 23(2)
• quantization noise is not so 5 large yet
• a restored digital signal after compression and decompression becomes as shown in FIG. 24( 1) , and hence large quantization noise is left.
• this restored digital signal is inputted into a variable low pass filter after the quantization bit length of this restored digital signal is lo expanded by predetermined bit length toward a low-order side, this restored digital signal becomes as shown in FIG. 24(2) , and hence the large quantization noise is left yet.
• an object of the present invention is to
• the present invention is characterized in that, in a method or an apparatus for processing a digital audio
• the present invention performs interpolation processing of signal levels within an interpolation object interval in accordance with a predetermined function curve, which monotonously changes, with the interpolation object interval, including a discontinuous part, which exists between one signal interval, where the same levels continue, and another signal interval, which is adjacent to the one signal interval and in which the same levels that are different from the former same level, continue.
• a digital signal that is an object of the interpolation processing is a series of quantized amplitude values of an audio signal in case of the digital audio signal, and a series of quantized gradation values or a series of quantized two-dimensional frequency conversion coefficient values of an image signal in case of a digital image signal.
• the interpolation processing of the present invention is performed by making a change characteristic of a function curve adaptively change in accordance with a degree of the periodicity of a signal in an interval covering intervals before and after the interpolation object interval. Then, the interpolation processing makes the characteristic of the function curve change so that the function curve may gently change if the degree of the periodicity of the signal is large, and the function curve may sharply change if the degree is small.
• the degree of the periodicity is detected by using frequency analysis or auto-correlation analysis.
• any one kind of interval is selected from among three kinds of intervals such as (i) an interval whose starting point is a start of a preceding same level continuation interval and whose end point is a start of a succeeding same level continuation interval, (ii) an interval whose starting point is an end of a preceding same level continuation interval and whose end point is an end of a succeeding same level continuation interval, and (iii) an interval whose starting point is a nearly median point of a preceding same level continuation interval and whose end point is a nearly median point of a succeeding same level continuation interval.
• the present invention makes gradation levels smoothly change along the horizontal direction by performing the interpolation processing in accordance with a predetermined function curve in a predetermined signal interval including a discontinuous part that exists between one signal interval, where the same gradation levels continue, and one signal interval, which is adjacent to the one signal interval and in which the same gradation levels that are different from the former continue, but also in the vertical direction of the frame screen, the present invention makes gradation levels smoothly change along the vertical direction by performing the interpolation processing in accordance with a predetermined function curve in a predetermined signal interval including a discontinuous part that exists between one signal interval, where the same gradation levels continue, and another signal interval, which is adjacent to the one signal interval and in which the same gradation levels that are different from the former continue. This corresponds to a case of processing a static image signal.
• the present invention makes gradation levels smoothly change along the interframe direction by performing the interpolation processing in accordance with a predetermined function curve in a predetermined signal interval including a discontinuous part that exists between one signal interval, where the same gradation levels continue, and another signal interval, which is adjacent to the one signal interval and in which the same gradation levels that are different from the former continue.
• the above-described method for processing a digital image signal includes a process of transforming a series of gradation levels in the horizontal direction into a series of gradation levels in the vertical direction after the step of performing the interpolation processing in the horizontal direction, a process of transforming a series of gradation levels in the vertical direction into a series of gradation levels in the interframe direction after the step of performing the interpolation processing in the vertical direction, and a process of transforming a series of gradation levels in the interframe direction into a series of gradation levels in the horizontal direction after the step of performing the interpolation processing in the interframe direction.
• the transformation of the series of gradation levels is performed by the reading/writing operation of frame memory, and preferably, with letting N be a natural number, the series of gradation levels is transformed by providing two image memories each having N frames of capacity and repeating reading and writing operation alternately.
• FIG . 1 is a block diagram showing a configuration of an interpolation apparatus according to an embodiment of the present invention
• FIG . 2 is a frequency spectral map showing a result of
• FIG. 3 is a flow chart showing the main processing of the DSP (Digital Signal Processor) in FIG. 1 ; l o FIG . 4 is a flow chart showing the main processing of the DSP in FIG. 1 ;
• FIG . 5 is a flow chart showing the FFT processing of the DSP in FIG. 1 ;
• FIG . 6 is a flow chart showing the data input i s processing of the DSP in FIG. 1 ;
• FIG. 7 is a flow chart showing the data output processing of the DSP in FIG. 1 ;
• FIG. 8 is explanatory diagrams showing the FFT operation of the DSP in FIG. 1 ; 20 FIG. 9 is diagrams showing an embodiment of interpolating operation by a curve;
• FIG. 10 is explanatory diagrams showing the interpolating operation of the DSP in FIG. 1 ;
• FIG . 1 1 is diagrams for explaining interpolation 25 functions
• FIG. 12 is diagrams showing interpolating operation by an interpolation apparatus
• FIG. 13 is a diagram showing interpolating operation by the interpolation apparatus in FIG. 1 ;
• FIG. 14 is a flow chart showing the main processing of the DSP according to an example modified from that in FIG. i ;
• FIG. 15 is a flow chart showing the main processing of the DSP according to another example modified from that l o in FIG. 1 ;
• FIG. 16 is diagrams showing another embodiment of interpolating operation by a curve
• FIG. 17 is a flow chart showing the main processing of the DSP according to still another example modified from 15 that in FIG. 1 ;
• FIG. 18 is a flow chart showing the main processing of the DSP according to a still further example modified from that in FIG. 1 ;
• FIG. 19 is diagrams showing an embodiment of 20 interpolating operation by a straight line
• FIG. 20 is a flow chart showing the main processing of the DSP according to another example modified from that in FIG. 1 ;
• FIG. 21 is diagrams showing another embodiment of 25 interpolating operation by a straight line
• FIG. 22 is a graph showing frequency characteristics of a conventional variable low pass filter
• FIG. 23 is diagrams for explaining the interpolating operation of a conventional variable low pass filter style
• 5 FIG. 24 is diagrams for explaining the interpolating operation of a conventional variable low pass filter style
• FIG. 25 is a graph showing the structure of image frames
• FIG. 26 is fundamental block diagrams showing the l o preliminary treatment and interpolation processing of a digital image signal
• FIG. 27 is conceptual diagrams of the transforming of a series of pixels in FIG. 26;
• FIG. 28 is a concrete circuit diagram of a 15 series-of-pixels transducer in FIG. 26 as an embodiment
• FIG. 29 is timing charts showing the operation of the circuit in FIG. 28.
• FIG. 30 is drawings showing a transversal filter executing the two-dimensional interpolation processing of 20 an image signal.
• the present invention adaptively performs the interpolation processing according to a waveform characteristic estimated.
• the present invention detects a degree of the periodicity, noise characteristics, impulse characteristics, or the like of the original waveform by analyzing a time interval (see FIG. 8) , which is widish and includes the intervals before and after a interpolation object interval (including the stepped part changing between the same value continuation intervals that adjoin to each other), by using frequency analysis and auto-correlation analysis.
• the present invention can reduce quantization noise with preventing some frequency component from being remarkably emphasized from the viewpoint of frequency components that an original series of digital data has before and after a first same value continuation interval by detecting a part where two same value continuation intervals, whose values are different from each other in the series of digital data, appear in succession, and performing 5 interpolation according to an interpolation function almost monotonously changing from the start of an interpolation object interval to the start of a second same value continuation interval with all or part of an interval, preceding just before the second same value continuation
• an interpolation waveform in the interpolation object interval By adapting an interpolation waveform in the interpolation object interval to the frequency components that a signal waveform of the series of digital data before the interpolation has in a predetermined interval also covering intervals before and after the first same value continuation interval, it becomes possible to reduce the quantization noise with preventing a frequency component from being especially emphasized from the viewpoint of 5 frequency components that the original series of digital data has.
• intervals before and after the first same value continuation interval means to perform processing lest needless frequency components should be preferably added by making respective spectrum frequencies of frequency components added by an i s interpolation waveform overlap, as much as possible, all or part of respective spectrum frequencies by the frequency component that the series of digital data before the interpolation has in the predetermined interval covering the intervals before and after the first same value continuation
• 25 interpolation has in the predetermined interval covering the intervals before and after the first same value continuation interval.
• an interpolation function is adaptively changed according to the relative amplitude of r a to RM .
• a normalized first function F t (x) is defined as follows:
• a normalized second function F 2 (x) is defined as follows: if R a ⁇ R M ,
• a function G(x) is defined as follows:
• this interpolation object interval is interpolated with the sinusoidal waveform with the frequency of f s / 2L ⁇ .
• the present invention preferentially performs reduction of the quantization noise if there is a small possibility of a frequency component being especially emphasized from the viewpoint of frequency components that the series of digital data before interpolation has before and after the first same value continuation interval even if quantization noise is extensively reduced, and performs the reduction of the quantization noise as far as a frequency component is not excessively emphasized if there is a large possibility of the frequency component being remarkably emphasized by attempting to extensively reduce the quantization noise.
• an output of an interpolation apparatus according to this embodiment is as shown in FIG. 12(2) , and hence it can be seen that the quantization noise is drastically reduced.
• an input series of digital data is as shown in FIG. 20( 1 )
• an output of the interpolation apparatus according to the present invention is as shown in FIG. 13 , and hence it can be seen that it is possible to suppress the quantization noise in comparison with a case of using a conventional low pass filter (see FIG. 24(2)) .
• the embodiment described above adaptively changes the interpolation function on the basis of the result of performing the frequency analysis (that is, the FFT analysis) of signals in the signal interval also covering intervals before and after the interpolation object interval
• analysis can be used instead of the FFT analysis, the analysis which is defined in the following equation and uses a short-time auto-correlation function.
• W(n) Window function
• S(n) Sampling signal
• the interpolation apparatus changes the interpolation function.
• the embodiment described above sets an interpolation object interval, whose length is the same as that of the second same value continuation interval, in the first same value continuation interval out of the first same value continuation interval and the second same value continuation interval, it can be performed in the contrary to set an interpolation object interval, whose length is the same as that of the first same value continuation interval, in the second same value continuation interval.
• the embodiment described above makes an interval as the interpolation object interval, the interval which is determined by making the start of the leading same value continuation interval as the starting point and making the start of the following same value continuation interval as the end point when there exist two same value continuation intervals adjacent to each other.
• an interval as the interpolation object interval the interval which is determined by making the end of the leading same value continuation interval as the starting point and making the end of the following same value continuation interval as the end point, and also in this case, the object desired can be achieved.
• Such a state that a signal level changes before and after the interpolation operation in this case is shown in FIG. 16.
• DSP digital signal processor
• FIG. 1 is a block diagram showing the structure of an interpolation apparatus embodying an interpolation method according to the present invention.
• Reference character 1 denotes an input terminal for digital data after decompression of compressed voice data, and here, it is assumed that a series of digital data do, d i , d 2 ) ... is inputted, the series of digital data which is sampled with a sampling frequency f s , has quantization bit length n, and is expressed in the notation of two's complement.
• DSP digital signal processor
• N and M are sufficiently large.
• Reference character 3 denotes a DSP, which receives a series of digital data do, d i, d 2 , ... from an input terminal 1 and writes the series in the first memory 2A with extending them by m bits toward the low-order side in an input order (concretely, in case of the notation of two's complement, m digits of O's are added to the low-order side of di if the MSB that is a sign binary digit of di is 0 , and m digits of l 's are added to the low-order side of di if the MSB that is the sign 5 binary digit of di is 1 ) .
• the DSP 3 performs predetermined interpolation processing of data stored in the first memory 2A to reduce the l o quantization noise.
• the DSP 3 reads data after interpolation from the first memory 2A in order in parallel to these writing to the first memory 2A and interpolation processing, and outputs the data from an output terminal 4 as a series of digital data Go, Gi, G2, ... , which is sampled
• FIGS. 3 and 4 are flow charts showing the main processing of the DSP 3
• FIG. 5 is a flow chart showing FFT
• FIG. 6 is a flow chart showing data input interruption handling by the DSP 3
• FIG. 7 is a flow chart showing data output interruption handling by the DSP 3.
• FIG. 8 is an explanatory diagram
• FIGS. 9 and 10 are explanatory diagrams showing an example of series of data stored in the first memory 2A, and hereinafter, interpolating operation will be described with reference to these drawings.
• data stored in an address i of the first memory 2A are expressed as Di, Di' , D(i) , Di, d (d: an arbitrary integer) , Di,d.
• QM 0 dB
• spectrum length of the amplitude value 0 dB 100 (see FIG. 2) .
• the DSP 3 not only clears all the regions of the first memory 2A and second memory 2B by means of initialization, but also makes a write pointer RP and a read pointer WP to the first memory 2A be 0 (step S 10 in FIG. 3) .
• the DSP 3 refers to the first memory 2A from the starting address side and searches whether an interval where a plurality of same value data continue exists in the input series of data (steps S l l and S 12 in FIG. 3) .
• the DSP 3 subsequently checks whether the end of the same value continuation interval is determined (step S 13) .
• the DSP 3 judges that the end is not determined if the newest input data corresponds to the end of the same value continuation interval found this time (see D 4 in FIG. 9 ( 1)) , and in this case, the DSP 3 waits for the end being determined with a subsequent data input. Differently from this, if data having a value different from that in the same value continuation interval is stored just after the same value continuation interval found this time (see D5 in FIG. 9(2)) , the DSP 3 judges that the end of the same value continuation interval is determined.
• the DSP 3 makes the same value continuation interval, found this time, as a first same value continuation interval, and stores its starting address, end address, data value, and a number of data as S i , E i , Y i , and L i respectively (step S 14) .
• S i 1
• E i 5
• Li 5
• the DSP 3 refers to the first memory 2A and checks whether a same value continuation interval having values different from those in the first same 5 value continuation interval exists just after the first same value continuation interval (step S 15) .
• the DSP 3 refers to the first memory 2A and searches a next same value continuation interval in the input series of data (steps S 16 and S 17) , and if the same value lo continuation interval is found, the process goes to the step S 13 for the DSP 3 to perform the same processing as described above.
• the DSP 3 subsequently checks whether the end of the same value continuation interval is determined (step S 18) .
• the DSP 3 judges that the end is not determined if the newest
• the DSP 3 judges that the end of the same value continuation interval is determined as the second same value continuation interval, and the DSP 3 stores its starting address, end address, data value, and a number of data as S2, E 2 , Y2, and L 2 respectively (step S 19) .
• S 2 6
• E 2 9
• L 2 4.
• the DSP 3 determines all of the first same value continuation interval or part of the first same value continuation interval just before the second same value continuation interval as an interpolation object interval, and hence the DSP 3 refers to the second memory 2B, checks whether FFT, including the end address E i of the first same value continuation interval in the object address range Ci to C 2 where the result of the FFT can be used, is completed, and if not, the DSP 3 waits for completion (step S20) .
• the DSP 3 performs the FFT processing, shown in FIG. 5, in parallel to main processing, shown in FIGS. 3 and 4, by using multitask processing.
• first with letting k, determining a range of data used for the FFT processing in the first memory 2A, be 0, and letting P, determining a memory region where the result of the FFT processing is stored in the second memory 2B , be 1 (step S3 1)
• the DSP 3 waits for the first to 512th data do, d i, ... , ds i i being inputted from the input terminal 1 and stored into the first memory 2A (step S32) .
• the DSP 3 performs the FFT processing by using these first to 5 12th data, obtains the amplitude of components (amplitude values and relative length of a spectrum; see FIG. 2) every frequency, and not only stores the amplitude in a first memory region of the second memory 2B, but also stores the address range C i to C2 , where the result of the FFT processing performed this time can be used, out of the first memory 2A in the first region (step S33) .
• step S33 the DSP 3 increments P to 2 (step S34).
• the DSP 3 increments k to 1 , and increments P to 3 (step S37) . Then, the DSP 3 waits for 5 the 513th to 1024th data being inputted from the input terminal 1 and stored into the first memory 2A (step S32) . If judgment is YES at the step S32 , the DSP 3 performs the FFT processing by using these 513th to 1024th data, obtains amplitude values every frequency and relative l o length of a spectrum, and not only stores them into a third memory region of the second memory 2B , but also stores the address range Ci to C2, where the result of the FFT processing obtained this time can be used, out of the first memory 2A in the third memory region (step S33) . As
• the DSP 3 repeats similar processing, analyzes frequency components by means of the FFT processing by using data in a certain period out of the
• the DSP 3 compares Li with L2
• step S2 1 in FIG. 4 determines a range, where L 2 data just before the second same value continuation interval out of the first same value continuation interval exists, as the interpolation object interval (step S22) . If Li ⁇ L 2 , the DSP 3 determines the entire first same value continuation interval as the interpolation object interval (step S27) .
• the DSP 3 checks whether r a / r ma ⁇ is equal to or more than a predetermined reference value Co (step S24) . Since the influence of an error caused by noise becomes large if r a / r max is less than Co, accurate interpolation cannot be performed, and hence the interpolation processing in the interpolation object interval found this time is not performed. If judgment is YES at the step S24, the DSP 3 performs interpolation processing in the interpolation object interval, found this time, by replacing values of digital data in the interpolation object interval in the first memory 2A according to an interpolation function whose values monotonously change from the start of the interpolation object interval to the start of the second same value continuation interval (step S25) . If judgment is YES at the step S2 1 and YES at the step
• the DSP 3 performs interpolation in the interpolation object interval found this time by replacing the data D(S ⁇ + j) in the address (S i + j) in the first memory 2A with Y(S ⁇ + j) (step S25) .
• D(3) , D(4) , and D(5) are replaced.
• step S25 the DSP 3 replaces S i with S 2 , E i with E 2 , Y i with Y2, and Li with L 2 , makes the second same value continuation interval, found this time, as a new first same value continuation interval (step S26) , and returns to step S 15 in FIG. 3.
• the DSP 3 refers to the first memory 2A, checks whether a same value continuation interval having values different from those in the first same value continuation interval, exists just after the first same value continuation interval, and if does not exist, the DSP 3 refers to the first memory 2A, and searches a next same value continuation interval in the input series of digital data (steps S 16 and S 17) , and if the same value continuation interval is found, the process goes to step S 13 for the DSP 3 to perform the same processing as described above.
• the DSP 3 subsequently checks whether the end of the other same value continuation interval is determined (step S 18) .
• the DSP 3 judges that the end is not determined if the newest input data corresponds to the end of the same value continuation interval, and the DSP 3 waits for the end being determined with a subsequent data input, but, if data having values different from those in the other same value continuation interval that is equal is stored just after the other same value continuation interval, the DSP 3 judges that the end of the other same value continuation interval is determined (see FIG.
• the DSP 3 compares Li and L 2 , and since Li ⁇ L 2 this time, the l o DSP 3 determines the entire first same value continuation interval as the interpolation object interval (step S27) . Since same value continuation intervals, which are similar to the first same value continuation interval, periodically appear frequently in the input series of digital data, the
• DSP 3 makes a sinusoidal wave component of an interpolation waveform in the interpolation object interval be suitable to a frequency component, which a signal waveform of the series of digital data before the interpolation has in a predetermined interval covering
• the DSP 3 refers to amplitude values every frequency stored in the first memory region including the address E i in the address range C i to C , where the result of frequency component analysis can be used with the FFT processing, out of the second memory 2B, lets an amplitude value in the frequency component with f s / 2L ⁇ be q a , lets relative length of a spectrum be r a , lets an amplitude value in the frequency f ma ⁇ , the value of which is the largest, be q ma ⁇ , and lets the relative length of its spectrum be r ma ⁇ (step S28) .
• the DSP 3 checks whether r a /r max is equal to or more than a predetermined reference value Co (step S29) . Since the influence of an error caused by noise becomes large if r a / r ma ⁇ is less than Co, the interpolation processing in the interpolation object interval found this time is not performed.
• the DSP 3 performs interpolation processing in the interpolation object interval, found this time, by replacing values of digital data within the interpolation object interval in the first memory 2A according to an interpolation function whose values are adaptively changed according to the result of the frequency component analysis with the FFT near the interpolation object interval to the series of digital data before the interpolation so that the values may monotonously change from the start of the interpolation object interval to the start of the second same value continuation interval (step S30).
• D(7), D(8), and D(9) are replaced.
• step S30 the DSP 3 replaces Si with S 2 , Ei with E 2 , Yi with Y 2 , and Li with L , makes the second same value continuation interval, found this time, as a new first same value continuation interval (step S26), returns to the step S15 in FIG. 3, and subsequently repeats the similar processing.
• the series of digital data Do, D i, D 2 , ... with quantization bit length n' bits after the interpolation that is stored in the first memory 2A is sequentially read at a period of l / f s with the data output interruption handling shown in FIG. 7 , and is outputted as the series of digital data Go , D i , D 2 , —
• the embodiment described above sets an interpolation object interval, having the length same as that of the second same value continuation interval, in the first same value continuation interval between the first same value continuation interval and the second same value continuation interval, as a modified example, it can be performed to set an interpolation object interval, having the length same as that of the first same value continuation interval, in the second same value continuation interval.
• the DSP 3 can perform the main processing shown in FIGS. 14 and 15 respectively instead of those in FIGS. 3 and 4 among the flow charts in FIGS. 3 to 7.
• the DSP 3 when performing the linear interpolation by setting an interpolation object interval in the first same 5 value continuation interval, the DSP 3 can perform the main processing shown in FIGS. 17 and 18 respectively instead of those in FIGS. 3 and 4 among the flow charts in FIGS. 3 to 7.
• FIGS. 17 and 18 Processing in FIGS. 17 and 18 will be simply described. l o Nevertheless, it is assumed that the input series of digital data do, d i, ... is the same as that in FIGS. 9 and 10. Quite similarly to steps S 10 to S 19 in FIG. 3 , data D i to D 5 in addresses 1 to 5 in the first memory 12A is made to be a first same value continuation interval, and addresses 6 to 9 i s are made to be a second same value continuation interval (steps S 10 to S 19 in FIG. 17; see FIG. 19( 1 )) .
• the DSP 3 compares Li with L (step S2 1 in FIG. 18) , and if Li > L2, the DSP 3 determines a range, where L2 data just before the second same value continuation
• the DSP 3 determines the entire first same value continuation interval as the interpolation object interval (step S27) .
• the DSP 3 since judgment becomes YES at step S2 1 , the DSP 3 performs interpolation processing by replacing values of digital data in the interpolation object interval in the first memory 2A according to an 5 interpolation function whose values linearly change monotonously from the start of the interpolation object interval to the start of the second same value continuation interval (step S60) after executing the processing at the step S22.
• the DSP 3 performs the interpolation processing by letting an interpolation function Y(S ⁇ + j) which indicates the relationship between an address of respective data in the interpolation object interval (S i + j) and the value Y of respective data after interpolation (j: 15 discrete variable; S i + j expresses an address in the first memory 2A) be:
• Y(S ⁇ + j) Y i + ⁇ (Y 2 - Y ⁇ ) / L 2 ⁇ .j
• j 1 , 2, ... , (L 2 - 1)
• D3, D 4 , and D5 in the addresses 3 , 4, and 5 in the first memory 20 2A according to the following formula:
• the DSP 3 replaces S i with S 2 , E i with E 2 , Y i with Y 2 , and Li with L 2 , makes the second same
• step S26 25 value continuation interval, found this time, as a new first same value continuation interval (step S26) , returns to the step S 15 in FIG. 17.
• the DSP 3 refers to the first memory 2A and checks whether a same value continuation interval having values different from those in 5 the first same value continuation interval exists just after the first same value continuation interval. If it does not exist, the DSP 3 refers to the first memory 2A and searches a next same value continuation interval in the input series of data (steps S 16 and S 17) , and if the same value l o continuation interval is found, the process goes to the step S 13 for the DSP 3 to perform the same processing as described above.
• the DSP 3 subsequently checks whether the end of the other same value continuation interval is determined (step S 18) . If
• the DSP 3 makes the other same value continuation interval as the second same value continuation interval, and stores its starting address, end address, data value, and a number of data as S 2 , E 2 , Y 2 , and L respectively (step S 19) .
• S 10
• E 2 Next, the process goes to step S21 in FIG. 18.
• the DSP 3 compares Li and L 2 , and since Li ⁇ L2, the DSP 3 determines all the first same value continuation interval as the interpolation object interval (step S27) .
• the DSP 3 can make a sinusoidal wave component of an interpolation waveform in the interpolation object interval be suitable to a frequency component, which a signal waveform of the original series of digital data before the interpolation has in a predetermined interval covering intervals before and after the first same value continuation interval by making the time length of the interpolation object interval be the same as that in the first same value continuation interval, and hence, it is possible to effectively reduce the quantization noise by preventing a frequency component from being excessively emphasized from the viewpoint of frequency components that the original series of digital data has before and after the first same value continuation interval.
• the DSP 3 performs the interpolation processing by replacing values of the digital data in the interpolation object interval in the first memory 2A according to an interpolation function whose values linearly change monotonously from the start of the interpolation object interval to the start of the second same value continuation interval (step S61).
• the DSP 3 replaces Si with S 2 , Ei with E2, Yi with Y 2 , and Li with L 2 , makes the second same value continuation interval, found this time, as a new first same value continuation interval (step S26), returns to the step S 15 in FIG. 17 , and subsequently repeats the similar processing.
• the series of digital data Do , D i , D2 , .. . with quantization bit length n' bits after the interpolation that is stored in the first memory 2A is sequentially read at a period of l / f s with the data output interruption handling shown in FIG. 7 , and is outputted as the series of digital data Go, D i, D 2 , ....
• FIGS. 17 and 18 perform linear interpolation by setting an interpolation object interval in the first same value continuation interval, it can be made to perform linear interpolation by setting an interpolation obj ect interval in the second same value continuation interval similarly to the case of the curve interpolation.
• the DSP 3 can perform the main processing shown in FIGS . 17 and 20 respectively instead of those in FIGS . 3 and 4 among the flow charts in FIGS. 3 to 7.
• the present invention can be similarly applied also in case a series of data do' , di' , d2' , ... after expanding do , d i , d2 , ... by m bits toward the low-order side 5 is inputted into the input terminal and the DSP writes these data into the first memory as it is. Processing of digital image signal
• an interpolation processing method of the present invention is l o performed with supposing a digital audio signal obtained by sampling and quantizing music sound or voice as a digital signal
• the interpolation method of the present invention can also be applied to a digital image signal.
• the interpolation processing of the digital image signal
• interpolation processing method of the present invention can also be applied to the digital image signal as it is, against the audio signal simply expressing time fluctuation of an instantaneous value of sound
• the image signal has such specificity that the image signal is accompanied by a two-dimensional visual spatiality.
• a TV signal is a time series of luminance levels (in the case of color, time series of respective R, G,
• reference character 25 is a diagram showing a frame concept as property accompanied with such an image signal
• reference character 10 denotes a current frame
• reference character 1 1 denotes a frame immediately after the current frame
• reference character I N denotes a frame after N frame time passing.
• plenty of small partitions in the current frame represent pixels, and for example, reference character 20 denotes the third pixel from the left on the tenth line.
• the horizontal direction (line direction) in a frame is called the x direction
• the vertical direction in the frame is the y direction
• the interframe direction is the z direction.
• An object of image signal interpolation processing of the present invention is a compressed digital image signal such as an inter-frame differential signal, or an in-frame digitization cosine function transformation (DCT) signal.
• DCT in-frame digitization cosine function transformation
• difference of gradation levels of respective pixels between frames is the object of the interpolation processing
• each 5 DCT coefficient obtained by blocking a frame and performing DCT transformation in each block is the object of the interpolation processing.
• these are simply called gradation levels and DCT coefficients.
• FIG. 26 is a diagram showing fundamental blocks for lo the interpolation processing of an image signal. It is assumed that, since the pixel signal includes gradation levels of three colors such as R, G, and B, first, this is separated, the interpolation processing is performed every color, these R, G, and B are synthesized after the i s interpolation processing, and an interpolated signal output is obtained (FIG. 26( 1)) .
• a pixel image signal is not composed of R, G, and B, but is composed of luminance (Y) and color difference (I, Q) , the pixel signal is separated into Y, I, and
• 25 levels or DCT coefficients is inputted into an x direction interpolation unit 30 , and the interpolation processing in the x direction (the horizontal direction in a frame) is performed.
• the series of levels in the x direction is transformed into a series of levels in the y direction by an x/y transformation unit 3 1 , and the interpolation processing in the y direction (the vertical direction in the frame) is performed.
• a series of levels in the z direction (the interframe direction) is obtained by a y/z transformation unit 33 , the interpolation processing in the z direction (the interframe direction) is performed, and finally, z/x transformation is performed, and the digital image signal, the interpolation processing of which is completed in all of the x, y, and z directions, is obtained as an output to be provided to an application device such as a display device.
• FIG. 27 is a conceptual drawing of series-of-levels transformation by means of this write/ read operation.
• FIG. 28 is a diagram showing a concrete circuit diagram as an embodiment of the series-of-levels transformation unit.
• reference characters 4 1 and 44 denote image memory with predetermined capacity
• reference characters 40 and 43 denote three-state buffers
• reference characters 42 , 45 , and 46 denote selectors that select one of inputs Xo and X i and make it as 5 a Y output
• reference character 49 denotes an address generator composed of a write address and a read address.
• reference character 47 denotes a signal generator generating a clock at a sampling frequency
• reference character 48 denotes a frequency divider. l o
• the transformation operation will be described with taking the x/y transformation processing as an example .
• the frequency divider 48 is made to divide the sampling frequency Sr by two times the frame of pixels (m x n) .
• the frequency Fr of the output F of the frequency i s divider is:
• the output F of the frequency divider is a signal where "L” and “H” are alternately switched every m x n times the sampling cycle (FIG. 29(d)).
• 20 41 and 44 are made to include capacity corresponding to a frame, that is, m x n locations whose addresses are expressed in the following matrix with m rows and n columns. (1,1) (1,2) (1,3) K (1, n)
• the horizontal writing and vertical reading of the image signal is controlled, and it is achieved to perform the x/y series-of-levels transformation that is the desired object.
• the y/ z transformation and z/x transformation with similar processing, in this case, it is necessary to provide memory capacity corresponding to intervals in the interframe direction to be interpolated.
• memory capacity corresponding to intervals in the interframe direction For example, in case processing in the interframe direction covering K frames is an object, it is necessary to provide image memory having capacity corresponding to K frames.
• in-frame 5 interpolation processing can be performed in a lump.

Landscapes

• Physics & Mathematics (AREA)
• Mathematical Physics (AREA)
• Engineering & Computer Science (AREA)
• Computer Hardware Design (AREA)
• Algebra (AREA)
• General Physics & Mathematics (AREA)
• Pure & Applied Mathematics (AREA)
• Compression Or Coding Systems Of Tv Signals (AREA)
• Complex Calculations (AREA)
• Compression, Expansion, Code Conversion, And Decoders (AREA)
• Editing Of Facsimile Originals (AREA)

Priority Applications (2)

Applications Claiming Priority (2)

Publications (1)

Family

ID=17726195

Family Applications (1)

Country Status (4)

Cited By (2)

Families Citing this family (8)

Citations (3)

Family Cites Families (3)

• 2000
  • 2000-04-06 JP JP2001530200A patent/JP4249414B2/ja not_active Expired - Fee Related
  • 2000-04-06 WO PCT/JP2000/002237 patent/WO2001028092A1/en not_active Ceased
  • 2000-04-06 US US10/089,939 patent/US6915319B1/en not_active Expired - Fee Related
  • 2000-04-08 TW TW89106511A patent/TW461196B/zh active

Patent Citations (3)

Also Published As

Similar Documents

Legal Events