JPH0548662B2 - - Google Patents

Description

[Detailed Description of the Invention] (Industrial Application Field) This invention relates to gamma correction processing of video signals required in video cameras, color correctors, etc., especially near low levels (black levels). This is related to signal processing.

(Conventional technology) Conventionally, when performing gamma correction using analog technology, the gain increases near the low level (black level), and random noise components included in the input signal here are amplified to a greater extent, resulting in dark The drawback was that noise in the scene became more noticeable and the overall image quality deteriorated. In order to eliminate this drawback, the gain at the black level should be set using an ideal gamma correction curve (where the input signal V _i and the output signal V _O are V _O =V〓 _i ,
For example, Υ is 0.45), and the gain has to be smaller than that, making it impossible to perform ideal gamma correction.

Another possible method is to perform gamma correction after band-limiting signals near the black level, but this method has the disadvantage that the resolution of the black level always deteriorates.

Furthermore, even when gamma correction is realized using digital technology, similar problems occur, and in order to deal with the increase in quantization errors due to the coarser quantization steps at lower levels, the video input signal must be quantized with sufficient precision in advance. expensive A/D
Required a converter.

(Problems to be Solved by the Invention) The present invention addresses the problems that occur near the black level when performing gamma correction on the video signal mentioned at the beginning, such as the fact that noise in dark scenes becomes more noticeable, and in digital processing, the lower the level, the more This is intended to solve the problem of increased conversion errors.

(Means for Solving the Problems) In order to solve the above-mentioned drawbacks, the present invention adds an auxiliary signal processing operation to the conventional gamma correction.

That is, in gamma correcting an input video signal, the gamma correction method of the present invention detects the input video signal level and the high frequency components of the input video signal, and detects that the input video signal is near the black level; In addition, when the high frequency component is small, the gamma correction is performed after band limiting the input video signal near the black level.

(Operation) FIG. 2 shows the input/output characteristics of a conventional gamma correction circuit. As mentioned above, the gain is high (slope) near the black level, so
A minute change Δl _i near the black level (which is usually considered to include more noise components and quantization error components in the input video signal) is greatly amplified by the output Δl _p . On the other hand, a minute change Δh _i near a high white level becomes Δh _p and is compressed.

The present invention attempts to reduce this increase in Δl _p , and for this purpose, signals at the black level are averaged between spatially adjacent pixels to improve the S/N ratio.

The averaging process is _o 〓 ^k=l v _i (t+k・Δt)/n (here
v _i is the input video signal voltage, n is the number of pixels to be averaged, and Δt is the sampling period between those pixels)
However, by effectively using the decimal part of the result obtained by the calculation without discarding it, an increase in quantization error in digital processing, for example, can be prevented. This processing is also suitable because quantization errors are more noticeable in areas where the pixel signal level changes gently.

Further, it also controls whether to stop or execute the averaging process depending on whether or not the signal level is changing rapidly.

In order to see changes in signal level, processing such as second-order differentiation is generally performed, but this degrades the S/N ratio. Therefore, the S/N before gamma correction
Detects level changes using an input signal that has not been degraded.

FIG. 1 is a block diagram of its principle circuit configuration. In FIG. 1, the video signal input (level _vi ) from the left side is guided to the gamma correction circuit 1, the averaging circuit 2, the contour detection circuit 3, and the level comparator 5 via the switch 7. The video signal input and the signal output from the averaging circuit 2 are gamma-corrected by the gamma correction circuit 1 selected by the switch 7 and output as a signal (v _p =v〓 _i ). This switching circuit is controlled by the contour signal comparison section (contour detection circuit 3
comparator 4) and signal level comparator 5. In other words, the switch 7 is selected for the averaging circuit 2 when the contour signal level (change in signal level or second-order differential) is smaller than a certain value (k ₁ ) and the input video signal level is below a certain value (k _{2 )} . ) is smaller than In other words, when the input video signal level is near the black level and there is no sudden change in the signal level, an averaged signal with a good S/N ratio is output.

The configuration of the gamma correction circuit 1 is not particularly special, and may be a broken line approximation method used in conventional analog technology, or a table lookup method using memory suitable for digital technology. Even signals near the black level (v _i <k ₂ ) are output directly from the gamma correction circuit 1 in portions where the contour signal level is greater than k ₁ .
In areas where the contour signal level is high, noise and quantization errors are less noticeable, so problems do not occur much.

As mentioned above, the averaging processing circuit 2 performs the calculation _o 〓 ^k=l _vi (t+k・Δt)/n...(1), but the positional relationship between the required pixel and the adjacent pixel is The weighting coefficient may be changed by _o 〓 ^k=l a _k · _vi (t+k·Δt)/m (2) (m is a constant).

For simplicity, the following will be explained using equation (1). Regarding the positional relationship of adjacent pixels, it is possible to use not only pixels on the right and left of a desired pixel, but also pixels located above, below, or diagonally. Now, it is assumed that the gamma correction value near the black level is approximated by v _p =l· _vi (3) (l is a constant). If the number of pixels to be averaged at this time is n, then v _p is calculated from equations (1) and (3) as follows: v _p =l・v _i =l・_o 〓 ^k=l v _i (t+k・Δt)/ n=l' _o 〓 ^k=l v _i (t+k・Δt)...(4) (Here, l'=l/n). If each constant is selected so that l=n, the output to the switch will simply be the sum of adjacent pixels, and the circuit will be simplified. The S/N improvement rate when doing this is 10log n (dB). If quantization error is not averaged, if the lowest bit (LSB) of the input video signal fluctuates, the output will fluctuate by n levels, making it impossible to express signal fluctuations below n levels. . However, by performing this averaging process,
Up to the LSB can be used effectively. This is because in equation (4), the calculation of l/n results in significant digits up to the LSB.

(Example) FIG. 3 shows a specific example in which l=n=8. Adjacent pixels of the main line signal are extracted by the delay element circuit 8. When performing averaging processing from 8 pixels on the top, bottom, diagonal, and left and right sides, this delay element circuit requires two 1-line delay lines and 6 1-pixel delay elements.
Realize by combining pieces. When 8 points of pixels are added in an averaging processing circuit, the S/N is improved by about 9 dB and output.

The contour detection signal is obtained by taking the difference between the output from the averaging processing circuit and the main line signal. This contour signal is input to the comparator 4 and compared with an appropriate constant _k1 , the change component of the signal due to noise (generally the signal level detected is small)
and signal change component due to contour (larger than above)
Distinguish them using k ₁ as the boundary.

On the other hand, the main line signal is also directly input to the comparator 5. In this comparator 5, as shown in FIG. 8b, the reference threshold level _k2 for selecting between the output from the averaging circuit and the direct video signal input is the object of comparison. In this way, the input/output characteristics when o < v _i < k ₂ are
v _p =8v _i . The quantization error is, for example, if an 8-bit video input is received, and the output is not averaged, the LSB change input will be 8 times the black level at that output, so from the LSB to the 3rd bit. This results in a dead zone, but if you perform averaging processing, you can output valid numbers down to the LSB.

In FIG. 3b, area A is the output area from the averaging circuit, and area B is the area for direct video signal input.

In the case of Fig. 3, only a linear approximation of the slope 8 was performed in the range of o<v _i <k ₂ , but furthermore, for example, if k ₂ <
In the range of v _i <k ₃ , it is also possible to perform broken line approximation near the black level, such as approximation to a straight line with a slope of 4.
In this case, one more comparator 5 is added and k ₂ <
When it is in the range of v _i <k ₃ , this information may be sent to the averaging circuit to control the switching from 8-point addition to 4-point addition.

(Effects of the Invention) By implementing the present invention, it is possible to prevent deterioration of image quality and increase of quantization error near the black level when performing gamma correction by performing spatial averaging processing. Therefore, effective gamma correction can be applied.

According to this method, a contour signal is also obtained at the same time, so the circuit size does not need to be so large for devices such as video cameras and image quality correctors that incorporate gamma correction circuits and contour correction circuits.

[Brief explanation of the drawing]

Fig. 1 is a block diagram of the principle circuit configuration for explaining the method of the present invention, Fig. 2 is a diagram showing the input/output characteristics of a conventional gamma correction circuit, and Fig. 3a is a concrete implementation of the method of the present invention. FIG. 3b, a block diagram of an example circuit, is a diagram showing the input/output characteristics at that time.

Claims (1)

[Scope of Claims] 1. When performing gamma correction on an input video signal, detecting the signal level of the input video signal and the high-frequency components of the input video signal, and detecting that the input video signal is near the black level; When the high-frequency component is small, the gamma correction method is characterized in that the gamma correction is performed after band-limiting the input video signal near the black level. 2 If the input video signal is a digital signal,
The gamma correction according to claim 1, wherein the band limitation is achieved by performing averaging processing on adjacent pixels, and the lower digit signal value generated by the averaging processing is also output. Method.