JPH0365047B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a digital filter applied to a low-pass filter of a digital color encoder.

When a digital filter is incorporated as hardware in a digital signal processing circuit as a real-time processor, it is usually used as a coefficient multiplier.
The method used is to look up a large capacity storage device such as ROM. As a result, the circuit scale increases, making it difficult to incorporate the digital filter into an LSI.

Therefore, the applicant of this application expanded the transfer function of a digital filter into a plurality of polynomials, set the coefficients of the plurality of polynomials to integers, made the coefficients correspond to the plurality of polynomials, and obtained the coefficients by addition. We have previously proposed a digital filter with a structure in which partial filters are connected in cascade.

By the way, in order to obtain the overall desired frequency characteristics (maximum flattening characteristic, Tiebishiev characteristic, etc.),
As the partial filter, a configuration is used in which a band rejection filter and an outside band boost filter are combined. In this case, as shown in FIG. 1, a configuration may be considered in which a band rejection type partial filter 3 and an out-of-band boost type partial filter 4, such as a high frequency boost type, are sequentially connected from the input terminal 1 to the output terminal 2. .

Now, the spectrum 5 shown by the solid line in Figure 2A
A signal having a spectrum 6 shown by a broken line
When an input signal containing quantization noise is supplied to the input terminal 1, signals with spectra 7 and 8 in FIG. 2B and quantization noise appear at the output of the band rejection filter 3. Even in the stop band, the quantization noise does not become 0, but this is because there is arithmetic quantization noise due to the output word length limit. Then, by passing through the high frequency boost type filter 4 having the frequency characteristics shown as 9 in FIG. As shown in spectrum 11, the quantization noise outside the band is enhanced.

When configuring a low-pass filter to limit the band of color difference signals provided in a digital color encoder that processes the output of a digital camera, it is necessary to avoid , as mentioned above, it is undesirable for out-of-band quantization noise to be enhanced.

In consideration of the above-mentioned points, the present invention is designed to connect a band rejection filter 3 after the high frequency boost filter 4, as shown in FIG.

Filter 4 having frequency characteristics 9 similar to those described above
An input signal having the spectrum shown in FIG. 4A is supplied to the input signal, and a signal having spectra 12 and 11 and quantization noise appear at the output, as shown in FIG. 4B. Then, it is supplied to the band rejection filter 3, and the 10
A signal with a spectrum shown by and quantization noise with a spectrum shown by 13 are generated. In this way, there is no difference in signal components between the configurations of FIG. 1 and FIG. 3. However, it is possible to prevent the quantization noise outside the band from being enhanced, and it is possible to suppress the residual noise up to the arithmetic quantization noise due to the output word length limit.

Below, this invention will be described as FIR (Finite Impulse Response)
An embodiment of the filter applied to a low-pass filter for band limiting a digital color difference signal (I signal) will be described with reference to FIG.

Input terminal 1 has input from the matrix circuit, for example.
An I signal (8-bit parallel) with a sampling frequency of 2f _SC (f _SC : color subcarrier frequency) is supplied. Further, D represents a unit delay amount of 1/2f _SC , and a unit delay circuit is realized by a latch circuit. Delay circuits 14, 18 and partial filters 15, 16, 17,
19 and 20 are connected in cascade. The delay circuits 14 and 18 have the same circuit configuration as the low-pass filter for the Q signal, and are inserted to make the group delay times equal.

The partial filters 15, 16, 19, and 20 are configured to add the input and output of the unit delay circuit, each having a transfer function of (Z+1), and are band rejection filters. In addition, the partial filter 17
The transfer function of is −Z ² +16Z ² −8(Z ³ +Z)+2(Z ³ +Z) +Z ⁴ +1=Z ⁴ −6Z ³ +14Z ² −6Z+1. This partial filter 17 is a high frequency boost type filter.

Note that the above-mentioned transfer function ignores the terms of group delay time and gain for the sake of simplicity.

The digital filter having the configuration shown in FIG. 5 has a low-pass characteristic with attenuation of -2 dB or less at 1.3 MHz and -20 dB or more at 3.6 MHz, and limits the band of the I signal to 1.5 MHz.

Further, the present invention can also be applied to a Q signal low-pass filter for limiting the Q signal band to 0.5 MHz. In that case, the transfer functions H ₁ (z) to _{H 7} (z) of the partial filters are shown below in the order from the input terminal to the output terminal.

H ₁ (z)=Z ² +1 H ₂ (z)=Z ³ +1 H ₃ (z)=Z ³ +1 H ₄ (z)=15Z ⁴ +8Z ³ +18Z ² +8Z+15 H ₅ (z)=-Z ³ +2Z ⁵ +2Z ⁴ +2Z ³ -1 H ₆ (z)=Z+1 H ₇ (z)=Z+1 The partial filters having the above transfer functions H ₄ (z), H ₆ (z), and H ₇ (z), respectively, The partial filter that is a blocking filter and has a transfer function H ₅ (z) is a high-frequency boost filter.

As understood from the above description, according to the present invention, out-of-band quantization noise can be suppressed to a level due to the output word length limit. Therefore, the word length between the stages of the partial filter can be determined by focusing only on the normal band, and the digital filter can be easily calculated.

Furthermore, if the configuration is such that partial filters whose transfer function coefficients are integers are connected in cascade as in the above embodiment, a large-scale data conversion table with an expanded data conversion table can be used.
It does not require ROM and can be implemented using random logic such as CMOS, making it suitable for LSI implementation. Furthermore, in order to prevent the scale of the ROM from becoming extremely large, the data conversion table itself is an approximation value, which causes adverse effects such as changes in characteristics. However, since the present invention is designed with coefficients as integers from the beginning, such problems do not occur.

Note that the present invention is not limited to low-pass filters, but can be similarly applied to high-pass filters or band-pass filters.

[Brief explanation of drawings]

1 and 2 are block diagrams and frequency characteristic diagrams used for reference in explanation of this invention, FIGS. 3 and 4 are block diagrams and frequency characteristic diagrams used in this invention, and FIG. 5 is a part of this invention. FIG. 2 is a block diagram showing the configuration of an embodiment. 1... Input terminal, 2... Output terminal, 3... Band rejection type filter, 4... High frequency boost type filter, 15, 16, 17, 19, 20... Partial filter.

Claims (1)

[Scope of Claims] 1 A plurality of partial filters corresponding to each of the polynomials given when a transfer function is expanded into a product of a plurality of polynomials are connected in cascade, and the plurality of partial filters include at least an out-of-band filter. In a digital filter that has a partial filter having a boost type characteristic and a band rejection type partial filter and having a flat passband characteristic, the band rejection type partial filter is provided after the partial filter having the out-of-band boost type characteristic. A digital filter with a partial filter connected to it.