JP2717688B2 - Imaging device - Google Patents

Description

Description: BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image pickup apparatus provided with a color image pickup means using a complementary color filter, and to a digital signal processing apparatus. Hereinafter, a video camera will be described as an example, but the present invention is also applicable to an electronic still camera or the like.

[Conventional technology]

Due to recent advances in digital signal processing technology, digital signal processing, which has advantages such as high reliability and no adjustment, has become mainstream in signal processing circuits of color video cameras. On the other hand, a color separation filter attached to the sensor often uses a complementary color rather than a pure color. This is because although pure colors have excellent color reproducibility, they have low light utilization and are disadvantageous in terms of S / N and bandwidth of luminance signals.

[Problems to be solved by the invention]

Among these two tendencies, a very general configuration example (conventional example 1) for performing digital signal processing in a single-panel color video camera using a complementary color filter is described in the seventh section.
Shown in the figure.

In the sensor 201, a complementary color filter as shown in FIG. 9 is formed for each pixel cell Ce, and performs interlaced scanning for each field in frame reading.

In this case, the signals of Mg and Gr are repeatedly read out during the first horizontal scanning period, and the signals of Ye and Cy are repeatedly read out during the second horizontal scanning period. The read signal is input to the A / D converter 211 via the AGC circuit 202 and the sample hold circuit 210. Now, assuming that the full scale of the A / D converter 211 is O to AmV and the conversion accuracy is 10 bits, each signal of Mg, Gr, Ye, and Cy has a random quantization noise (error) of Δ / 2 only. Are considered to be superimposed.

here It is.

The luminance signal is basically the read signal Mg, Gr, Ye, C
Since it is obtained by processing y, the order of the magnitude of the quantization noise in the luminance signal is approximately Δ / 2. This luminance signal is subjected to γ correction and the like in the luminance signal processing circuit 205, and then the D / A converter 20
The signal is input to the NTSC encoder 209 via 7.

On the other hand, the color signal is subjected to color processing in the color signal processing circuit 206 after the output of the A / D converter 211 is once converted into a color difference form such as (Mg-Gr) or (Cy-Ye) by the subtractor 212. . In other words, the color signal component is modulated into a luminance signal component, and the ratio of the color signal component to the normal luminance signal component is 20 to 30%.
It is as follows. This means that the color signal processing circuit 206
Then, in order to obtain a sufficient level of a color signal, the original signal must be multiplied by 3 to 5 times. Therefore, the magnitude of the quantization noise in the chrominance signal is three to five times that of the luminance signal. Of the order. Thereafter, the output of the color signal processing circuit 206 is combined with the luminance signal by the NTSC encoder 209 via the D / A converter 208 to form an NTSC signal. Therefore, in such a configuration, the number of bits of the A / D converter 211 must be at least 12 bits in order to secure the final S / N ratio of the color signal, which increases costs and increases power consumption. Was invited.

To solve this problem, as shown in Fig. 8, A / D
Two converters are provided, one of which takes a difference in advance by a differential amplifier 313, adjusts a gain, and performs A / D conversion. The result is used as a color difference signal.
A method of forming a luminance signal by performing A / D conversion in the A / D converter 315 at the full scale corresponding to the output sampled and held in step 1 (conventional example 2) is also conceivable. Note that, for example, when the read clock is fc, the sample and hold circuits 310 and 311 have a cycle of 2 / fc and the phases are inverted from each other. In this configuration, the deterioration of the S / N ratio of the color signal due to the quantization noise is slightly improved, but is still insufficient.

The present invention has been made under such circumstances,
It is an object of the present invention to provide an imaging device in which quantization noise of a color signal is inconspicuous.

[Means for solving the problem]

The solution of the present invention for achieving the above object is based on the following recognition of a and b.

a. In the case of a normal subject, the color signal difference, that is, the histogram of the color difference signal is close to a normal distribution centered on 0.

b. In human visual characteristics, where the level of the chrominance signal is small, that is, near the achromatic color, even a slight error of the chrominance signal is detected, but where the level of the chrominance signal is large, that is, the color saturation Is sufficiently high, there is a characteristic that even a small error in the color difference signal cannot be detected.

Therefore, based on the above two recognitions, the present invention
The smaller the level of the color difference signal, the finer the quantization, and the higher the level of the color difference signal, the coarser the quantization.

More specifically, the present invention relates to the following (1) and (2)
It is configured as follows.

(1) Color imaging means using a complementary color filter, separation means for receiving the output of the color imaging means and separating each complementary color signal, and differential amplifier means for inputting a plurality of complementary color signals separated by the separation means Conversion means for non-linearly A / D-converting the output of the differential amplifying means, and processing means for linearizing the output of the conversion means, so as to finely quantize small signal levels. I do.

(2) In the above (1), it is assumed that the conversion means includes a resistive voltage dividing circuit for generating a divided voltage output having a partially different value, and a plurality of comparators using the divided voltage output as a reference value.

[Action]

According to the above configurations (1) and (2), a portion where the level of the color difference signal is small is quantized finely, and a portion where the level of the color difference signal is large is roughly quantized.

〔Example〕

 Hereinafter, the present invention will be described with reference to examples.

FIG. 1 is a block diagram of a video camera according to a first embodiment of the present invention.

In the figure, a sensor 101 is a CCD image sensor, on which a complementary color filter shown in FIG.
With 103, reading is performed by interlaced scanning similar to the conventional example.

The gain of the read signal is adjusted by the AGC circuit 102, and the read signal is input to the two sample and hold circuits 110 and 111. Pulses P1 and P2 are supplied from the timing generation circuit 104 to the sample and hold circuits 110 and 111, respectively. As shown by the circles in FIGS. 2 (b) and 2 (c), P1 and P2 are synchronized with the read clock fc of the sensor at a cycle twice as long as this.
And the phases are mutually inverted.

FIGS. 2 (a), (b), (c) and (d) show Mg.
Each AGC circuit in horizontal scanning period reading out / Gr line
FIG. 3 shows a schematic diagram of an output of 102, outputs of sample and hold circuits 110 and 111, and an output of a differential amplifier 112. FIG. One scale is time T,
T = 1 / fc. In the figures (b) and (c), the timings of the pulses P1 and P2 are indicated by circles, and the timing of the pulse at which the A / D converter 114 operates is shown in (e).

Thus, for example, in the first horizontal scanning period, the outputs of the sample and hold circuits 110 and 111 become Mg and Gr, respectively, and in the second horizontal scanning period, they become Ye and Cy, respectively. Therefore, the output of the differential amplifier 112 connected to the sample and hold circuits 110 and 111 is given by Is repeated every 1H (horizontal scanning period).

However, in the case of FIG. 9, the arrangement of the color filters in the first field and the second field is reversed, so that P1, P
The two pulses must be phase inverted with respect to each other between the first and second fields.

Now, let the maximum value of the output of the sample hold circuit 110 and the sample hold circuit 111 be V (mV). Differential amplifier 1
The possible values of the twelve outputs are in the range -KV to KV mV.

However, actually, when examining the histogram of the output at the point (refer to the upper center in FIG. 1), the color is different, but the difference between horizontally adjacent signals has a horizontal correlation of luminance.
It is almost in the range of -αKV to αKV mV. α is smaller than 1 and approximately 1/5 to 2/5.

Also, from the viewpoint of human visual characteristics, the error is conspicuous when the level of the color difference is small. Accordingly, the compression characteristic as shown in FIG.
It should be realized in 7.

In this embodiment, α = 2/5 β = 4/5, and when | x | <αKV, slope 2 when αKV <| x | <βKV, slope 1/2 | x |> βKV, clipping is performed. It has become.

However, the conversion range of the A / D converter 114 is -A / 2mV to A / 2
In the case of mV. This can be easily implemented using a circuit as shown in FIG. The first bending point ± αKV is approximately ± 0.
Set to 7V. The second bending point ± βKV is βKV = V ₂ +
0.7V, -βKV = - may be set as (V ₃ + 0.7V). Further, the inclination of the first described bending points ± ArufaKV to the second described bending points ± BetaKV be determined by the ratio of R ₁ and R _in.

The A / D converter 114 is an 8-bit linear A / D converter having a conversion range of -A / 2mV to A / 2mV, and inputs an input in the range of -A / 2 to -A / 2 + [Delta] to 0,-. The input in the range of (A / 2−Δ) to (A / 2) mV is converted into 255 steps. The linear ratio processing unit 118 includes the A / D converter 1
Reverse conversion of the conversion performed by the nonlinear processing unit 117 is performed on the digital output of 14. This can be done by writing the above-described inverse conversion function in the ROM in advance and outputting data when input digital data is used as an address.

On the other hand, the outputs of the sample and hold circuits 110 and 111 are
Entered into 13. If the gain here is L Are alternately output from the adder 113 every 1H. Therefore, the conversion range of the A / D converter 115 is 0 to 2 LVmV
It is good to set to.

If the A / D converter 115 has a precision of 10 bits or more, the A / D converter 115 may be used as it is.
Before the converter 115, it is preferable to provide a non-linear compression circuit in the same way as for the color difference signal, and then to return to linear.

The output of the A / D converter 115 is input to the luminance signal processing 125, where processing such as low-pass filter, γ conversion, Knee conversion, edge enhancement, and luminance step correction is performed, and the final luminance signal Y is obtained. .

For improving color reproducibility, it may be corrected only reduce part of the Y with the reduced correction signal Y _L of Y obtained in the manner described below.

 FIG. 5 shows the configuration of the two-dimensional interpolation filter 121.

The output of the linearization processing unit 118 is a 1H memory 501 and a constant multiplier 50.
Entered in 3. For example, when the input of the 1H memory 501 is a signal corresponding to P, the input of the 1H memory 502 is Q, and the output is P
Is a signal corresponding to.

Therefore, if the constants of the constant multipliers 503 and 505 are set to 1/2 and the constant of the constant multiplier 504 is set to 1, the outputs of the adders 506 and 504 have P and Q alternately every 1H and different from each other. Appears in phase. Therefore, these are input to the switch 507, and this and 1
If switching every H, point is always P, point is always Q
Are obtained, and P and Q are simultaneously synchronized and simultaneously low-pass filtered in the horizontal direction. The color band is considered to be sufficient if it is 160 to 240 TV lines in the vertical direction.Therefore, by using two or more 1H memories or configuring an IIR type filter to narrow the band in the vertical direction, It is desirable to reduce the occurrence of color moiré.

The horizontal low-pass filters 508 and 509 are digital low-pass filters in the horizontal direction, respectively, and narrow the band to about 1 to 2 MHz. This may be either FIR type / IIR type.

In this way, after sufficiently lowering the band, the thinning units 510, 51
1 reduces the number of data to N: 1. Clock is f _C , bandwidth is Then, N is preferably selected as an integer slightly smaller than f _C / f _D. Thereby, ready to perform subsequent matrix processing and color processing operation a relatively low speed clock f _D, it is effective in order to reduce power consumption.

On the other hand, the output of the A / D converter 115 is also a two-dimensional interpolation filter 121.
Is input to the two-dimensional interpolation filter 120 having the same configuration as the above, so that the output of 120 becomes X and Y signals interpolated two-dimensionally.

Next, the scaling unit 128 will be described. For example, A
The A / D converter 115 has 8 bits, and the A / D converter 114 has 8 bits.
= 4, the LSB of the output of the A / D converter 115 is And the LSB of the output of the linearization process 118 is The scale of the latter is 8 times finer. Therefore, scaling conversion is performed such that the former is uniformly moved to the left by 3 bits and becomes 11 bits.

 Next, the matrix calculation unit 122 will be described.

After scaling, it is considered that X = Mg + Gr Y = Cy + Ye P = Mg−Gr Q = Cy−Ye It is.

Now, by the matrix (a _ij ) of A (4 × 3), Mg, G
It is assumed that r, Cy, Ye can be converted to R, G, B.

The coefficient (a _ij ) may be determined as follows.

That is, the spectral characteristics of the color filters of each color are represented by Mg (λ), Gr
(Λ), Cy (λ), Ye (λ) R (λ), G (λ), B (λ) are, for example, NTSC
It is preferable to determine the ideal imaging characteristic by using the least square method or the like.

Using (a _ij ) determined in this way, Becomes

Such an operation may be performed once for the clock of f _C / N.

In order to prevent a false color from occurring in the vertical direction, the coefficient (a _ij ) preferably satisfies the following condition.

It is. Here, U and V are constants.

Therefore, (6) is Becomes

In equation (8), for example, since the calculation unit of UX + VY is common in the calculation of each color of R, B, and B, if this is calculated first, and then the other part is calculated, the calculation circuit and calculation time Can be reduced.

Next, the R, G, and B outputs of the matrix operation unit 122 are subjected to white balance processing, γ conversion, and color difference matrix processing in a color signal processing unit 123, and color difference signals RY and BY are output. You.

The luminance signal Y is converted by a D / A converter 126 into color difference signals RY and BY.
Are converted from digital to analog by a D / A converter 124 and input to a standard television signal generator 127, and then become standard television signals NTSC.

 Next, a second embodiment of the present invention will be described with reference to FIG.

This embodiment is different from the configuration shown in FIG.
An A / D converter having a configuration as shown in FIG. 6 is used instead of the part 7 and the A / D converter 114.

The resistor voltage dividing circuit 91 has a total of 256 resistors R _i (i: 1 to 25
6). For example, the resistance value of the resistor R ₂₅₆ to R ₂₀₀ and the resistor R ₁ to R ₅₇ 4R, resistors R ₅₆ to R ₁₂₈ and the resistor R ₁₂₉
The resistance value of R199 to _R199 may be R. Also, R ₂₅₆ and the resistor R ₁ is connected to an external reference power supply ± βKV (mV). Further, between the resistors R ₁₂₉ and the resistor R ₁₂₈ is grounded. C
_{i (i: 1~256)} is a comparator, which comparator the upper voltage and the input voltage of the R _i. These 256 results are input to the encoder 92, and the results are encoded into 8 bits.

By using the A / D converter having such a configuration, the third
It is possible to obtain the characteristics as shown in FIG. 7B, and the other configuration may be the same as that of the first embodiment.

〔The invention's effect〕

As described above, according to the present invention, the output of the differential amplifying means is non-linearly A / D-converted and then linearized, so that the color difference signal is fine when the level is small and coarse when the level is large. It is quantized, and color quantization noise (error) can be made inconspicuous to human visual perception, so that an imaging device with high quality image quality can be provided.

[Brief description of the drawings]

FIG. 1 is a block diagram of the first embodiment, FIG. 2 is a schematic diagram of the output of each unit of the first embodiment, FIG. 3 is an explanatory diagram of the non-linear processing unit 117, FIG. 5 is a block diagram of the two-dimensional interpolation filter 121, FIG. 6 is a circuit diagram of the second embodiment, FIG. 7 is a block diagram of the conventional example 1, FIG. 8 is a block diagram of the conventional example 2, and FIG. FIG. 7 is a layout diagram of complementary color filters. 101 Sensor 110, 111 Sample hold circuit 112 Differential amplifier 114 A / D converter 117 Non-linear processing unit 118 Linear processing unit

Claims (2)

(57) [Claims] 1. A color image pickup means using a complementary color filter,
A separating unit that receives the output of the color imaging unit and separates each complementary color signal; a differential amplifying unit that receives a plurality of complementary color signals separated by the separating unit as input; An imaging apparatus comprising: a conversion unit for performing D-conversion; and a processing unit for linearizing an output of the conversion unit, and quantizes a portion having a small signal level more finely. 2. The converter according to claim 1, wherein said converting means includes a resistive voltage dividing circuit for generating a divided voltage output having a partially different value, and a plurality of comparators using said divided voltage output as a reference value. An imaging device according to any one of the preceding claims.