JPH03136586A - Compression coder - Google Patents

Abstract

PURPOSE:To efficiently eliminate noise generated in an inter-frame prediction encoding loop by applying ON/OFF control to a bank limit filter so as to apply band limit to an inter-frame prediction decoding signal when lots of picture elements with a larger amplitude difference (distortion) exist between an input signal and a decoding signal. CONSTITUTION:A subtractor 2 calculates an inter-frame prediction error, a prediction error signal 2a is subject to orthogonal conversion by an orthogonal conversion circuit 3, quantized at a quantizer 4 and converted into a format to be sent to a multiplex circuit 14 and sent to a compression decoder. An arithmetic circuit 9 calculates a difference absolute value between an input television signal 1a and an inter-frame prediction decoding signal 6a obtained through decoding a coded signal, outputs a selection signal 9a to apply on/off control of a spatial filter 11. Noise is eliminated from a signal distorted by orthogonal conversion and quantization by passing it through a band pass filter in the spatial filter 11.

Description

[Detailed description of the invention] [Industrial application field] The present invention relates to a compression encoding apparatus for television signals.

[Conventional technology]

As a compression coding technology for transmitting video television signals over low-speed transmission channels, only moving parts (prediction error signals) are processed using interframe predictive coding (also known as interframe predictive coding with motion compensation). There is technology to transmit. Furthermore, in order to compress information, orthogonal transform coding such as cosine transform is used for the prediction error signal generated by interframe predictive coding, and hybrid coding is performed in which only transform components with large power are coded and transmitted as significant components. technology, and has achieved significant information compression in video television signals. However, the high frequency components transformed by orthogonal transformation generally have low power and may not be transmitted.

Therefore, when interframe predictive coding is used, high frequency components remain in the predictive coding loop and are output as noise in the decoded signal.

In order to solve the above problem, conventional compression encoding devices use a band-limiting filter (a filter that spatially cuts high frequency components: a filter that spatially cuts high frequency components) for the predictive decoded signal, which is the interframe predicted value generation source, in the predictive encoding loop. (This is why it is called an in-loop filter.) A band-limiting filter is applied to the moving parts to cut out high-frequency components. Furthermore, to detect a moving part, a motion vector detected during motion-compensated interframe predictive coding was used.

[Problem to be solved by the invention]

The conventional compression encoding device described above requires a motion detection circuit because it is necessary to detect a moving part for on/off control of the in-loop filter. Further, depending on errors in motion detection or characteristics of the in-loop filter, noise may not be completely removed and may remain in the loop.

[Means for Solving the Problems] The compression encoding device of the present invention performs compression encoding in which a prediction error signal obtained by interframe predictive encoding of a television signal is orthogonally transformed encoded, and the transformed components are quantized and transmitted. In the encoding device, a filter for band-limiting the interframe predictive decoded signal, a first means for determining an absolute difference value between the input television signal and the interframe predictive decoded signal, and a first means for determining the absolute difference value between the input television signal and the interframe predictive decoded signal, and a second means for calculating the number of pixels exceeding the first threshold for each predetermined block; and a second threshold in which the number of pixels exceeding the first threshold is predetermined. a third means for selecting the filter pressure for each block when the filter pressure exceeds the predetermined value; and selecting the interframe predictive decoded signal for each block when the filter pressure does not exceed the predetermined value; It has a fourth means for generating a predicted value, and a fifth means for transmitting a selection signal for controlling the third means for each of the blocks.

〔Example〕

Next, the present invention will be explained with reference to the drawings.

FIG. 1 is a block diagram showing one embodiment of the present invention.

An input television signal 1a is input to an input terminal 1, which is connected to a subtracter 2 and a delay circuit 7. A prediction error signal is calculated in the subtracter 2, the prediction error signal 2a is connected to an orthogonal transform circuit 3, a transform component 3a is quantized in a quantizer 4, and the quantized transform component 4a is sent to a multiplexing circuit 14. It is connected to the inverse orthogonal transform circuit 5 as well. The prediction error signal 5a inversely transformed by the inverse orthogonal transform circuit 5 is added with a predicted value 8a by an adder 6 to produce an interframe predictive decoded signal 6.
a as an arithmetic circuit 9. Connected to delay circuit IO and spatial filter 11. The interframe predictive decoded signal 10a delayed by the delay circuit 10 and the filtered interframe predictive decoded signal 11a are connected to a selection circuit 12, and the selected decoded signal 12a is connected to a predicted value generation circuit 13.

The generated predicted value 13a is connected to the subtracter 2 and the delay circuit 8, and the predicted value 8a delayed by the delay circuit 8 is connected to the adder 6. The PCM television signal 7a delayed by the delay circuit 7 is connected to an arithmetic circuit 9, and a selection signal 9a is output from the arithmetic circuit 9, which is connected to a selection circuit 12 and a multiplexing circuit 14. A multiplexed signal 14a is outputted from the multiplexing circuit 14 to an output terminal 15.

Next, the specific operation will be explained. The subtracter 2 calculates an interframe prediction error, and performs a difference calculation with the previous frame. The calculated prediction error signal 2a is orthogonally transformed in an orthogonal transform circuit 3. An example often used in orthogonal transform encoding of image signals is discrete
There is cosine transform coding (DCT). DCT converts an image signal on the time axis into a signal on the frequency axis, and divides the prediction error signal 2a into blocks each consisting of predetermined n lines and Xm pixels, and converts a numerical sequence into a cosine function for each block. This is an orthogonal transformation using . In image signals, power is concentrated in the low frequency components of the cosine-transformed components, so it is common to suppress the amount of generated information by quantizing the low frequency components finely and the high frequency components coarsely. is being carried out. Therefore, each component subjected to orthogonal transformation is quantized by the quantizer 4,
The data is converted into a format for transmission by the multiplexing circuit 14 and transmitted to the compression/decoding device. The multiplexing circuit 14 variable-length encodes the quantized signal using a Huffman code or the like to configure a frame for transmission, and also multiplexes a selection signal 9a and other mode codes, which will be described later.

On the other hand, in order to use the quantized transform component 4a as a generation source of an interframe predicted value signal, the inverse transform of the orthogonal transform circuit 3 is performed in the inverse orthogonal transform circuit 5, and the signal on the time axis is converted from the signal on the frequency axis. signal (prediction error signal), and adder 6
By adding the predicted value 8a, the television signal is decoded as an interframe predictive decoded signal 6a.

The spatial filter 11 is a filter that removes noise by passing a signal distorted by orthogonal transformation and quantization through a band-limiting filter, and generally there is a lot of noise in high-frequency components with low power, which is visually harmful. Therefore, a low-pass filter is used to remove high frequency components. In addition, since orthogonal transformation is a two-dimensional orthogonal transformation for high-efficiency encoding, the filter is also spatial (two-dimensional).
It is common to use a filter.

The delay circuit 10 is a circuit for compensating for the delay in the spatial filter 11.

The selection circuit 12 selects a filtered signal 11a or an unfiltered signal 10a using the selection signal 9a.
Either one is selected and supplied to the predicted value generation circuit 13.

In the case of simple inter-frame predictive coding, the predicted value generation circuit 13 uses a frame memory to serve as a delay memory for supplying the subtracter 2 with a signal of one frame before, and in the case of motion-compensated inter-frame predictive coding, This is a variable delay memory whose delay changes depending on the motion vector signal.

Delay circuit 7 and delay circuit 8 are delay compensation memories for adjusting signal delays.

The arithmetic circuit 9 is a circuit for generating a selection signal 9a that controls the selection circuit 12, and generates the absolute value of the difference between the input television signal 1a and the interframe predictive decoded signal 6a obtained by decoding the encoded signal. and calculate the number of pixels whose absolute difference value exceeds a predetermined first threshold value for each predetermined block;
If the number of pixels showing a value exceeding the first threshold value in each block exceeds the second threshold value, signal 11a is applied to the signal of the block; otherwise, signal 10a is applied.
A selection signal 9a is output so as to select each of them.

The block serving as the on/off control unit of the spatial filter 11 is generally a block for orthogonal transformation (n lines x
m pixels), and the generation of distortion due to quantization of orthogonal transform components of each block is handled. for example,
If the orthogonal transform block is 8 lines x 8 pixels, the filter control block is also 8 lines x 8 pixels, and is in one-to-one correspondence with the orthogonal transform coding block.

A block diagram of an example of the arithmetic circuit 9 is shown in FIG.

In FIG. 2, a subtracter 901 calculates a difference between an input television signal 7a and an interframe predictive decoded signal 6a to calculate a difference value 901a, which is converted into an absolute value 902a in an absolute value conversion circuit 902. Conversion to absolute value is R
This can be achieved by code conversion using OM. Absolute difference value 902
a is determined to be larger than the threshold value 1 by the comparison period 903,
Depending on whether the pixel exceeds the threshold l (effective pixel) or not, it is converted into a code of “1” or “O” and is supplied to the scan conversion circuit 904.

In order to perform filter control on a block-by-block basis, it is necessary to
) is input to the scan conversion circuit 90.
4, the signal is converted into a block scanning signal as shown in FIG. 3(b). Scan-converted valid/invalid display signal 904a
A counting circuit 905 counts the number of effective pixels for each block.

The count value of the counting circuit 905 is reset for each block by a reset signal. The effective coefficient value 905a per block counted by the counting circuit 905 is compared in magnitude with a threshold value 2 in a comparator 906, and when it exceeds the threshold value 2, the output signal 11a of the spatial filter 11 is selected. Similarly, a selection signal 906a is generated. selection signal 90
6a is converted from block scanning to TV scanning by a scan conversion circuit 907, and is sent to the selection circuit 12 as a selection signal 9a.
is supplied to

〔Effect of the invention〕

As explained above, the present invention provides a filter that limits the band of the interframe predictive decoded signal, calculates the absolute difference value between the PCM television signal and the interframe predictive decoded signal in predetermined block units, and calculates the absolute value of the difference between the input signal and the decoded signal. By controlling the band-limiting filter on and off to limit the band of the inter-frame predictive decoded signal when there are many pixels with large amplitude differences (distortion) from the signal, the quantization of orthogonal transform coding It is possible to efficiently remove noise generated in the predictive coding loop, and also to be able to control the band-limiting filter in the loop without providing a motion detection circuit as in the conventional case.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing one embodiment of the present invention, FIG. 2 is a block diagram showing an example of the arithmetic circuit 9 of FIG. 1, and FIG.
Figures a) and (b) are diagrams for explaining block scanning. 1...Input terminal, 2...Subtractor, 3.
...Orthogonal transform circuit, 4...Quantizer, 5
...Inverse orthogonal transform circuit, 6...Adder,
7, 8, 10...Delay circuit, 9...Arithmetic circuit, 11...Spatial filter, 12...
... Selection circuit, 13 ... Predicted value generation circuit, 14
...Multiplex circuit, 15...Output terminal,
901...Subtractor, 902...Absolute value conversion circuit, 903, 906...Comparator, 904
, 907... Scan conversion circuit, 905...
・Counting circuit.

Claims (1)

[Claims] A compression encoding device that orthogonally transform encodes a prediction error signal obtained by interframe predictive encoding of a television signal, quantizes and transmits the transformed component, and a filter that band-limits the interframe predictive decoded signal; a first means for calculating an absolute difference value between an input television signal and the interframe predictive decoded signal; and a block unit in which the number of pixels for which the absolute difference value exceeds a predetermined first threshold is predetermined. a second means for calculating the first
When the number of pixels exceeding a threshold exceeds a predetermined second threshold, the filter output is selected for each block, and when the number does not exceed a second threshold, the interframe predictive decoded signal is selected for each block. a fourth means for generating an inter-frame predicted value from the signal selected by the third means; and a fifth means for transmitting a selection signal for controlling the third means for each block. A compression encoding device characterized by having means.