JP2678066B2 - Digital signal processor - Google Patents

Description

Description: BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a digital signal processing device, and more particularly to a digital signal processing device suitable for encoding and decoding digital image signals.

[Prior Art] For the conventional encoding and decoding of digital image signals, refer to, for example, the document "Multidimensional signal processing of TV images" (Takahiko Fukibe, published by Nikkan Kogyo Shimbun, published November 15, 1988). As shown, an encoding method by offset sampling and a decoding method by data interpolation are known. FIG. 4 shows a block diagram of such a conventional digital signal processing apparatus. In FIG. 4, (1) is for inputting a digital video signal obtained by digitizing an analog image signal such as a television signal. A data input terminal, (21) a prefilter for band-limiting the input signal from the data input terminal (1), and (27) a predetermined rule for a signal band-limited by the prefilter (21). The re-sampling circuit (12) for re-sampling based on the above is a data output terminal for sending the signal encoded as described above to a transmission / recording system (30) for transmitting or recording data. The data input terminal (1), the prefilter (21), the resampling circuit (27), and the data output terminal (12) constitute a coding block (51) of the digital signal processing device. The data from the transmission / recording system (30) is sent to the decoding block (52), (14) is the data input terminal of the decoding block (52), and (24) is the resample of the coding block (51). Circuit (2
The data interpolation circuits (24) and (25) that interpolate the missing portion of the data sampled in 7) add the data interpolated by the data interpolation circuit (24) to the sampling data input from the data input terminal (14). A selector for switching data for interpolation insertion, and (19) is a data output terminal for outputting a digital video signal decoded and reproduced by the decoding block (52).

The operation of the above arrangement will be described below with reference to the explanatory view of FIG. By the way, FIG. 5 shows the sampling points of the video signal two-dimensionally by the circles and the X marks.

When the digital video signal input from the data input terminal (1) is digitized from the analog video signal, it is sampled at the points indicated by the circles and the X marks in FIG. This input data is band-limited by the pre-filter (21) and re-sampled by the resampling circuit (27). This resampling is performed in the continuous video data as shown in the explanatory diagram of FIG. Is discarded by discarding the data marked with X and making the data marked with ◯ valid (so-called sub-sampling). In this case, offset sampling is performed by shifting the sampling point for each scanning line of the image. The data rate transmitted from the data output terminal (12) of the coding block (51) to the transmission / recording system (30) by the above sub-sampling is 2 minutes of the data rate input from the data input terminal (1). Will be 1.

The transmission / recording system (30) interposed between the coding block (51) and the decoding block (52) by the above coding.
For transmission, the system needs only half the bandwidth, and if the bandwidth of the system is the same, the transmission speed doubles. On the other hand, in recording, the number of recorded images can be doubled.

The decoding block (52) is used to restore the data from the transmission / recording system (30) to the original video signal.
Data from the recording system (30) is input to the data input terminal (14). Since this data includes only the portion marked with a circle in FIG. 5, it is necessary to reproduce the data of the X mark thinned out by the resampling circuit (27) of the coding block (51). . The method used to reproduce this data is interpolation, and this interpolation operation is performed, for example, by using the data D (y _{m + 1, n} ) at y _{m + 1, n} points in FIG. 5 and the data D at y _{m, n} points before and after it.
(Y _{m, n} ) and the linear interpolation D (y _{m + 1, n} ) = {D (y _{m, n} ) as shown in the equation (1) of predicting from the data D (y _{m + 2, n} ) of _{ym + 2, n} points _n ) + D (ym _{+ 2, n} )} / 2 (1) is performed. This interpolation calculation is carried out by the data interpolation circuit (24), and the data indicated by X in FIG. 5 is reproduced.
Since the data marked with a circle in FIG. 5 is input from the data input terminal (14), the selector (25) selects the data marked with a circle from the data input terminal (14) and the X from the data interpolation circuit (24). Alternately select the data marked with and output the data (19)
By sending the data to the digital input terminal (1), a decoded digital video signal approximate to the digital video signal input from the data input terminal (1) can be obtained.

[Problems to be Solved by the Invention] Since the conventional digital signal processing device is configured as described above, the data of the X-marked pixels that do not actually pass through the transmission / recording system (30) are the circled-marked pixels on both sides. The predicted value is interpolated by the average value of the data. Therefore, there is a problem that an interpolation error, that is, a prediction error becomes large in a portion where the video signal has a wide band such as an edge portion of the image, and the image based on the decoded video signal is deteriorated.

The present invention has been made to solve the above problems, and it is a digital signal processing capable of reducing a prediction error in an edge portion of an image or the like and reducing deterioration of an image based on a decoded video signal. The purpose is to obtain the device.

[Means for Solving the Problem] In order to achieve the above object, the present invention is a sampling means for thinning out n-1 pieces of n pieces of digital video data to sample and transmit only one piece of data. And thinned n
A plurality of prediction means for predictively encoding one piece of data by a plurality of different prediction methods, and the prediction signal in each of the prediction means and the thinned n-1 pieces of data are matched to detect each prediction error. And a means for transmitting the judgment signal of the prediction method that minimizes the prediction error and the minimum prediction error signal by comparing the detection errors of the prediction error detection means with the sampling data. Interpolation means for generating n-1 interpolation data based on the sampled data, the prediction method and the minimum prediction error signal, and n digital images based on the transmitted sampling data and the interpolation data from the interpolation means. The present invention provides a digital signal processing device having means for decoding data. Further, the determination signal is characterized by instructing a prediction method that minimizes a value calculated based on the n-1 prediction errors. Note that the value calculated based on n-1 prediction errors means one value obtained by comprehensively considering n-1 prediction errors using an arbitrary function, and n-1 The maximum value of the absolute value of the prediction error, the variation value of the n-1 prediction errors, or any function value such as the root mean square error is included.

[Operation] In the above means, in the digital signal processing device of the present invention, the sampling means thins out n-1 pieces of the n pieces of digital video data to sample and transmit only one piece of data, and a plurality of predictions. N-thinned out by means
One piece of data is predictively encoded by a plurality of different predicting methods, and the predictive error detecting means detects the predictive error in each predictive method by matching the predictive signal in each predictive means with the thinned n-1 data. Then, the detection signals are compared with each other and the discrimination signal of the prediction method that minimizes the prediction error and the minimum prediction error signal are transmitted together with the sampling data, and the sampling data transmitted by the interpolating means and the prediction method and the minimum are transmitted. N-1 interpolation data are generated based on the prediction error signal, and n digital image data are decoded based on the transmitted sampling data and the interpolation data from the interpolation means. Further, as the discrimination signal, n-1
Since the prediction method that minimizes the value calculated based on the prediction error is specified, one determination signal is sufficient for n-1 prediction errors.

Hereinafter, embodiments of the present invention will be described with reference to the drawings.

FIG. 1 is a block diagram of a digital signal processing apparatus according to an embodiment of the present invention. In the figure, (2) is a memory for holding the required amount of digital video signals input from the data input terminal (1), and (3) is required from the memory (2) based on a command from the control circuit (13). Data sending circuit for selecting and sending various data, (10) is a first quantizer for quantizing the data sent from the data sending circuit (3), and (4) is a first quantizer (10) A first predictor for predicting interpolation data from the output data of
(5) is the output data from the first quantizer (10) and the first
Prediction error detector and second prediction device (6) for calculating a difference in prediction data from the prediction device (4) to obtain a prediction error
Is a second predictor that predicts the interpolated data from the output data of the first quantizer (10) by a method different from that of the first predictor (4), and (7) is the first quantizer ( A second prediction error detector for calculating a prediction error by calculating a difference between the output data from 10) and the prediction data from the second predictor (6), and (8) a first prediction error detector (5 ) And a prediction error from the second prediction error detector (7), (9)
Is a second quantizer for quantizing the prediction error signal A sent from the comparator (8), and (11) is data from the first quantizer (10) and the second quantizer (9). ) And a predictor discriminating signal B sent from the comparator, which is selected based on a command from the control circuit (13) and sent to the data output terminal (12). . The coding block (51) is composed of the system from the data input terminal (1) to the data output terminal (12).

The data at the data output terminal (12) of the encoding block (51) is the transmission / recording system (3
0), but the data from this transmission / recording system (30) is then sent to the decoding block (52).

In the decoding block (52), (15) is a memory that holds the data necessary for decoding from the data input from the data input terminal (14), and (16) is a memory from the control circuit (13). A data transmission circuit that selects and transmits necessary data from the memory (15) by a command, (38) is a third predictor that predicts interpolation data based on the data from the data transmission circuit (16), and (39) is data Sending circuit (1
4) Predict interpolated data based on data from 6)
Of the data transmitting circuit (16), the third predictor (38), and the fourth predictor (39), and (18) the data transmitting circuit (16). ) Output and adder (1
It is a selector that alternately selects the output of 7) to reproduce the digital video signal and sends it to the data output terminal (19).
Above data input terminal (14) to data output terminal (19)
The decoding block (52) is configured in the system up to.

By the way, the third predictor (3
8), the fourth predictor (39) is configured to perform exactly the same data prediction as the first predictor (4) and the second predictor (6) of the coding block (51).

The operation of the above-described structure will be described below with reference to the explanatory diagrams of FIGS.

In this embodiment, for example, encoding as shown in FIG. 3 is performed. In other words, assuming the data compression rate is 1/2,
Considering the data for 4 pixels as one set, 1 data out of 4 data is sampled and simply requantized to 6 bits, and the remaining 3 data transmit a prediction error of 3 bits. Here, since the digital video data is represented by 8 bits, if the 4 data represented by 32 bits can be coded and represented by 16 bits, the compression rate can be halved. In this embodiment, since the digital video signal can be encoded with 15 bits, the remaining 1 bit is used as the predictive encoding method determination signal.

Further, FIG. 2 shows a conceptual diagram when a digital image signal at the time of encoding is transmitted on the screen. In the same figure, the ∘ mark is data in 6-bit representation, the X mark is prediction error data in 3-bit representation, and represents a part of one frame image.

By the way, the digital video signal input from the data input terminal (1) has the data amount necessary for prediction stored in the memory (2). The data transmission circuit (3) is controlled by the control circuit (13) and stores necessary data in the memory (2).
Take out from and send out. The data sent from the data sending circuit (3) is quantized by the first quantizer (10) into, for example, 6-bit data. Here, if the data currently handled corresponds to the circles in FIGS. 2 and 3, the data is sent to the selector (11) and output from the data output terminal (12). If the data currently handled corresponds to the X mark, the data is predicted by the first predictor (4) and the second predictor (6), and the first predictive error detector (5), the prediction error is calculated by the second prediction error detector (7), the prediction method with the smallest error is selected by the comparator (8), the prediction method with the smallest error is selected, and the prediction error is the largest. A signal indicating a prediction method with a small number of predictors, that is, a predictor determination signal B and a prediction error signal A for determining an optimum predictor are transmitted. Here, in the comparison in the comparator (8), 4 data is 1
Since it is considered as a unit, the prediction method with the smallest prediction error is determined after comprehensively considering the prediction errors of the three data, and is transmitted as the predictor discrimination signal B. Along with that, the prediction error signal A is set as a group of 3 data
Input to the quantizer (9) and re-quantized into 3-bit representation data. The selector (11) is a control circuit (1
Under the control of 3), the signal of the first quantizer (10) selects the signal of the second quantizer (9) or the signal from the comparator (8), and 16-bit data is set to 1 The encoded signal is output from the data output terminal (12) as a delimiter.
In the present embodiment, in the first predictor (4) and the second predictor (6), for example _, x _{m, n + 1} , x in FIG.
Pixel data of X mark of _{m, n + 2} and x _{m, n + 3} is predicted as follows. First, in the first predictor (4), the prediction in the x direction is And the prediction in the y direction is made by the second predictor (6), Becomes The prediction method with the smallest prediction error is determined as follows, for example. The absolute value of the prediction error generated by the first predictor (4) is expressed as follows.

(E _{m, n + 1} ) _I = | (x _{m, n + 1} ) _I− x _{m, n + 1} | (8) (E _{m, n + 2} ) _I = | (x _{m, n + 2} ) _I− x _{m, n + 2} |・ ・ ・ (9) (E _{m, n + 3} ) _I = | (x _{m, n + 3} ) _I− x _{m, n + 3} │ ・ (10) The one with the largest absolute value of these prediction errors Max ((E _{m, n + 1} ) _I , (E _{m, n + 2} ) _I , (E _{m, n + 3} ) _I } ... (11). Similarly, the absolute value of the prediction error generated by the second predictor (6) is expressed as follows.

( _{Em, n + 1} ) _II = | (xm _{, n + 1} ) _II- xm _{, n + 1} | ... (12) ( _{Em, n + 2} ) _II = | (xm _{, n + 2} ) _II- xm _{, n + 2} |・ (13) (E _{m, n + 3} ) _I = | (x _{m, n + 3} ) _II −x _{m, n + 3} | ・ (14) The maximum absolute value of these prediction errors is calculated as Max. {(E _{m, n + 1} ) _II , (E _{m, n + 2} ) _II , (E _{m, n + 3} ) _II } ... (15). Therefore, Max {(E _{m, n + 1} ) _I , (E _{m, n + 2} ) _I , (E _{m, n + 3} ) _I } -Max {(E _{m, n + 1} ) _II , (E _{m, n + 2} ) _II , (E _{m , n + 3} ) _II } ··· (16) If the calculation result is a negative value, the prediction method by the first predictor (4) is the prediction method with the smallest prediction error. The prediction method by the second predictor (6) is determined to be the prediction method with the smallest prediction error.

On the other hand, the data input terminal (14) of the decoding block (52)
The digital signal to be decoded, which is input from, holds the data required for decoding in the memory (15). The data transmission circuit (16) is controlled by the control circuit (13), takes out the data necessary for decoding from the memory (15), and transmits the data to the block designated by the control circuit (13). That is, the data corresponding to the circles in FIGS. 2 and 3 are sent to the selector (18) and X
The data corresponding to the mark is the third based on the predictor discrimination signal B.
Of the second predictor (39) or the fourth predictor (39). When the data is predicted by the third predictor (38) or the fourth predictor (39), the predictive error signal A sent from the data sending circuit (16) in the adder (17). Is added and sent to the selector (18). Here, as described above, 4 data is 1
Since it is considered as a unit, as shown in FIG. 2 or FIG. 3, the data of the pixel of the X mark in which three data are continuous is made up of one set of three data and the predictor discrimination signal B
A third predictor (38) or a fourth predictor (39) according to
Data of each X is not sent to different predictors.

The selector (18) is controlled by the control circuit (13) and alternately selects the signal from the data transmission circuit (16) or the adder (17) with four data as one set. as a result,
The decoded digital video signal is output from the data output terminal (19).

In the above embodiment, the prediction interpolation of the data is performed by using the vertical linear prediction and the horizontal linear prediction as the data prediction method, but other prediction methods such as polynomial prediction may be used. An oblique condition may be entered as the prediction direction. Further, not only one-dimensional prediction but also two-dimensional prediction such as plane prediction, and three-dimensional prediction based on data between several fields and several frames may be used.

Further, in the above embodiment, the case where four data are considered as one group as shown in FIG. 3 is illustrated, but an arbitrary number of data may be considered as one group, and the number of bits expressing a digital video signal or The number of bits expressing the prediction error signal can be set freely.

Furthermore, in the above embodiment, in determining the prediction method that minimizes the prediction error, the configuration of comparing the magnitude of the absolute value of the prediction error is illustrated, but the magnitude of the variation range of the prediction error is compared, and the prediction error It may be determined by a method such as comparing the magnitude of the root mean square error.

[Effects of the Invention] As described above, according to the present invention, data is predicted by a plurality of encoding methods, and a prediction error when data is predicted by a prediction method that has the smallest prediction error is transmitted. Since it is configured as described above, there is an effect that a prediction error is minimized and a deterioration in image quality due to a decoded digital video signal can be minimized.

[Brief description of the drawings]

FIG. 1 is a block diagram of a digital signal processing apparatus according to an embodiment of the present invention, FIGS. 2 and 3 are explanatory views for explaining the operation of the configuration of FIG. 1, and FIG. FIG. 5 is a block diagram of the signal processing device, and FIG. 5 is an explanatory diagram for explaining the operation of the configuration of FIG. (1) is a data input terminal, (2) is a memory, (3) is a data transmission circuit, (4) is a first predictor,
(5) is a first prediction error detector, (6) is a second predictor, (7) is a second prediction error detector, (8) is a comparator,
(9) is the second quantizer, (10) is the first quantizer,
(11) is a selector, (12) is a data output terminal, (13) is a control circuit, (14) is a data input terminal, (15) is a memory, (16) is a data sending circuit, (17) is an adder,
(18) is a selector, (19) is a data output terminal, (30) is a transmission / recording system, (38) is a third predictor, (39) is a fourth predictor, and (51) is a coding block. , (52) is a decoding block. In the drawings, the same reference numerals indicate the same or corresponding parts.

Claims (2)

(57) [Claims] 1. n-1 of n digital video data
Sampling means for thinning out and transmitting thinned pieces, and thinned out n-1
A plurality of predicting means for predictively encoding each piece of data by different predicting methods, a predictive error detecting means for detecting each predictive error in each predicting means, and a predictive error by comparing each detective error of the predictive error detecting means. A means for transmitting the discriminant signal of the prediction method and the minimum predictive error signal that minimize the error together with the sampled data, and n based on the transmitted discriminant signal of the sampled data and the predictive error and the minimum predictive error signal.
Based on the transmitted sampling data and the interpolating data from the interpolating means,
A digital signal processing apparatus comprising means for decoding individual digital image data. 2. The digital signal processing apparatus according to claim 1, wherein the discrimination signal indicates a prediction method that minimizes a value calculated based on the n-1 prediction errors.