JP2010021694A - Edge enhancement image processing apparatus - Google Patents

Abstract

An edge-enhanced image processing apparatus that can perform accurate and effective edge enhancement without the need to provide a special table.
A luminance signal is generated from an input RGB image signal, and gamma correction is performed. The gamma correction table used for gamma correction is the same as that used for gamma correction of the main image signal, but the accuracy of the output signal can be output with higher accuracy than that required by the main image signal. Shall. Then, edge detection and edge adjustment are performed to generate an edge signal and rounded to the accuracy of the main image signal, and then the edge signal is added to the luminance signal of the main image signal.
[Selection] Figure 1

Description

  The present invention relates to an edge enhanced image processing apparatus that performs edge enhancement of a camera image.

  Today, electronic cameras have become widespread and generally used widely, but further improvement in image quality of electronic cameras is desired. In order to improve the image quality, edge enhancement is performed to improve the appearance of the image. A conventional edge enhancement method will be described below.

FIG. 8 is a diagram for explaining a first conventional edge-enhanced image processing apparatus.
An RGB signal, which is a raw image signal, is input and γ correction is performed. At the time of γ correction, for example, an image signal that was initially a 12-bit signal is converted into an 8-bit low-accuracy image signal. In this way, the accuracy of the image signal is lowered in order to match the display monitor standard and the image data format such as JPEG or MPEG.

  After YC conversion of the γ-corrected image signal with a reduced bit length, edge detection is performed from luminance Y, the detected edge component E is adjusted, and then edge enhancement E is added to luminance Y. Do.

  The problem of the first prior art is that the bit length of the luminance Y is small (8 bits) and the accuracy is too rough for edge detection.

FIG. 9 is a diagram for explaining a second prior art edge-enhanced image processing apparatus.
As shown in FIG. 9A, the luminance Yl is generated from the RGB image, the edge is detected from the luminance Yl, the detected edge component E is adjusted, and then the edge component Ead is added to the RGB image. To perform edge enhancement.
This method is described in Patent Document 1.

  A problem of the second prior art is that the edge-enhanced effect becomes difficult to see due to gamma correction in a bright portion where the luminance Y is high. That is, even if edge enhancement is performed before gamma correction, the effect is crushed after gamma correction, making it difficult to see. This is illustrated in FIG. 9B. The γ correction has a curve characteristic as shown in FIG. When the luminance before correction is large, the slope of the curve is small, and therefore the amount of change in luminance before correction is small after correction. Therefore, even if edge enhancement is performed before γ correction and a change in luminance is given to the edge portion, the change in luminance is reduced after γ correction, and the effect of edge enhancement is reduced. .

FIG. 10 is a diagram for explaining a third conventional edge-enhanced image processing apparatus.
In FIG. 10, luminance Yl is generated from the RGB image, edge detection is performed from the luminance Yl, and the detected edge component E is adjusted. Thereafter, edge γ processing is performed in consideration of the gradient of gamma correction, and edge enhancement is performed by adding the edge component Eadg to the luminance Y after YC conversion of the main image signal.

This technique is described in Patent Documents 2 and 3. The edge γ processing here is to perform the following calculation.
(Edge γ rear edge component Eadg) = (Edge γ front edge component Ead) × (Edge γ gain)
The edge γ gain is obtained by referring to a preset table by luminance Y.
The edge γ corresponds to the case where the gradation of the luminance Y component is deformed by gamma correction.

A problem of the third prior art is that it requires table data for edge γ gain, and it is difficult to adjust edge γ.
JP 2006-333316 A JP 9-46554 A JP-A-2005-260517

The problem of the first prior art is that the bit length of the luminance Y is small (8 bits) and the accuracy is too rough for edge detection.
A problem of the second prior art is that the edge-enhanced effect becomes difficult to see due to gamma correction in a bright portion where the luminance Y is high.

A problem of the third prior art is that table data for edge γ gain is required, and it is difficult to adjust edge γ.
An object of the present invention is to provide an edge-enhanced image processing apparatus that can perform accurate and effective edge enhancement without requiring a special table.

  The edge-enhanced image processing apparatus according to the present invention includes a first gamma correction unit that performs gamma correction on an input image signal using a gamma correction table that outputs an output signal having a first accuracy, Using a correction table, a second gamma correction unit that performs gamma correction at a second accuracy that is lower than the first accuracy, and an image signal using an output signal from the first gamma correction unit Edge extraction means for extracting the edge component of the second gamma correction means, an output of the second gamma correction means, and an addition means for adding the output of the edge extraction means.

  According to the present invention, it is possible to provide an edge-enhanced image processing apparatus that does not require a special table and can perform effective and effective edge enhancement.

The main process of camera image processing is
FIG. 1 is a diagram for explaining an edge-enhanced image processing apparatus according to an embodiment of the present invention.
The gamma correction unit 10 performs gamma correction on the RGB image, and then the YC conversion unit 11 converts the RGB image into a YCbCr image. Here, when gamma correction is performed, normal bit length compression (eg, 12 bits → 8 bits) is performed.

  In the edge emphasis of this embodiment, the Y generation unit 12 generates the luminance Yl from the RGB image, and the gamma correction unit 13 performs the same gamma correction as that of the main image signal. However, compression of bit length is not performed here (example: 12bit → 12bit). Edge detection is performed from the luminance Yg after gamma correction, the detected edge component E is adjusted, and then edge enhancement is performed by adding the edge component Ead to the luminance Y after YC conversion of the main image signal.

  Here, the image space before gamma correction is a linear space. The linear space is a space in which the pixel value increases in proportion to the energy of light from the subject. On the other hand, the image space after the gamma correction is a curved space. The curved space is a space in which the pixel value is not proportional to the energy of light from the subject and becomes large.

That is, the luminance Yl is generated from RGB before gamma correction. Gamma correction is applied to the luminance Yl. However, this gamma correction is 12bit → Abit (8 <A). This means that it is more accurate than RGB after gamma correction. However, the upper 8 bits of the output value of the gamma correction coincide with the output value of the gamma correction for RGB of the main image signal. The two gamma corrections differ only in accuracy.
Edge detection is performed on the luminance Yg after gamma correction to obtain an edge component E. Since the edge (unevenness) included in the image is detected, the accuracy of the image at this point is important. Next, the edge component E is adjusted. The contents of adjustment are as follows.
・ Suppression of noise component contained in edge component E.
・ Strength of edge enhancement.
-Adjust the accuracy to match the 8-bit accuracy of Y after YC conversion.

The adjusted edge component Ead is added to the luminance Y after YC conversion of the main image signal to obtain the edge-enhanced luminance Yee.
The operation of this embodiment will be described in more detail with reference to FIGS.

FIG. 2 shows a conversion graph of a common gamma correction table held in the gamma correction units 10 and 13. FIG. 3 is a diagram illustrating a Laplacian filter.
Gamma correction is performed on R, G, and B for the main image signal.
Rg = (GTbl [R] >> 4)
Gg = (GTbl [G] >> 4)
Bg = (GTbl [B] >> 4)
Here, GTbl is a gamma correction table, and the same table is used on the main image signal side and the edge enhancement side. As shown in FIG. 2, the gamma correction table GTbl used in this embodiment converts a 12-bit pre-correction signal into a 12-bit post-correction signal. However, on the main image signal side, the corrected signal uses only the upper 8 bits of the gamma correction table GTbl.
Next, YC conversion is performed on Rg, Gg, and Bg.

Y00〜Y22は、あらかじめ定められた係数である。

Y00 to Y22 are predetermined coefficients.

On the edge enhancement processing side, luminance Yl is generated from R, G, and B before gamma correction. An expression for obtaining the luminance Yl is, for example, as follows.
Yl = kr * R + kg * G + kb * B
For example, kr = 0.3 kg = 0.59 kb = 0.11
Gamma correction is performed on the luminance Yl (Yg = GTbl [Yl]: This equation indicates that the gamma correction table GTbl is indexed using the luminance Yl).

Edge detection is performed on the luminance Yg after gamma correction to obtain an edge component E.
E = Laplacian filter [luminance Yg]
The Laplacian filter is, for example, a filter as shown in FIG. Here, c = 4 × (a + b), for example, a = 1.0 and b = 1.0. The Laplacian filter is 3x3, but of course it can be larger. When a Laplacian filter is applied to an image, an edge component of the image is obtained.

Edge component E is adjusted to create edge component Eadl. The method for adjusting the edge component E is, for example, the following processing. This suppresses the noise component included in the edge component E.
If + Cor <E, Ead1 = E-Cor
-Cor ≤ E ≤ + Cor Ead1 = 0
If E <-Cor Ead1 = E + Cor
Here, Cor is a coring threshold parameter. Cor may be changed according to luminance Yg or luminance Y.

Further, for example, the following processing is performed. This is an adjustment of the edge strength.
Ead = Ead1 * Scl
Here, Scl is a scale parameter. Scl may be changed according to the luminance Yg or the luminance Y. Further, the accuracy of the edge enhancement signal is adjusted so as to match the 8-bit accuracy of luminance Y after YC conversion of the main image signal. In this case, an 8-bit signal is used.

The adjusted edge component Ead is added to the luminance Y after YC conversion of the main image signal by the adder 16 to obtain the edge-enhanced luminance Yee.
Yee = Y + Ead
Yee clips to fit in the range [0-255].

FIG. 4 is a diagram showing a configuration for tone control (tone correction).
The luminance Yl is generated by the Y generator 12a from R, G, B before gamma correction. The luminance Yl is generated according to the following equation.
Yl = kr * R + kg * G + kb * B
For example, kr = 0.3 kg = 0.59 kb = 0.11

The tone control gain generation unit 20 obtains the tone control gain β with the luminance Y1 by referring to the tone control gain table TcTbl. Then, tone control is performed on R, G, and B as shown in the following equation. That is, the multiplier 17 performs the following multiplication.
Rt = β * R
Gt = β * G
Bt = β * B

From this equation, the following equation holds.
Yt = kr * Rt + kg * Gt + kb * Bt
Yt = β * Yl
Rt, Gt, and Bt are clipped to fall within the range [0 to 4095].

  If the configuration of FIG. 4 is provided before the input RGB signal of FIG. 1 is branched to the Y generator 12, tone control can be performed on the entire image, including edge enhancement.

5 and 6 are diagrams illustrating a second embodiment including tone control.
In FIG. 5, the same components as those in FIGS. 1 and 4 are denoted by the same reference numerals.
From the R, G, and B before the gamma correction, the Y generator 12 generates the luminance Yl.
Yl = kr * R + kg * G + kb * B
For example, kr = 0.3 kg = 0.59 kb = 0.11
The tone control gain generation unit 20 obtains the tone control gain β with the luminance Y1 by referring to the tone control gain table TcTbl. Then, tone control is performed for R, G, B, and Y. That is, the multiplier 17 performs the following multiplication.
Rt = β * R
Gt = β * G
Bt = β * B
Yt = β * Yl
Rt, Gt, Bt, and Yt are clipped to be within the range [0 to 4095].

The gamma correction unit 10 performs gamma correction on Rt, Gt, and Bt.
Rg = (GTbl [Rt] >> 4)
Gg = (GTbl [Gt] >> 4)
Bg = (GTbl [Bt] >> 4)
Here, only the upper 8 bits of the gamma correction table GTbl are used.

Y00〜Y22は、あらかじめ定められた係数である。 Performs YC conversion for Rg, Gg, and Bg. YC conversion is performed according to the following equation.

Y00 to Y22 are predetermined coefficients.

For the luminance Yt, gamma correction is performed with reference to the gamma correction table. Here, the number of bits of the converted signal is 12 bits which is the same as the input.
Yg = GTbl [Yt]
Edge detection is performed on the luminance Yg after the gamma correction to obtain an edge component E.
E = Laplacian filter [Luminance Yg]
The Laplacian filter is the one described in FIG. 3, for example.

Next, edge adjustment is performed. The edge adjustment is, for example, the following process. This suppresses the noise component included in the edge component E.
If + Cor <E, Ead1 = E-Cor
-Cor ≤ E ≤ + Cor Ead1 = 0
If E <-Cor Ead1 = E + Cor
Here, Cor is a coring threshold parameter. Cor may be changed according to luminance Yg or luminance Y.

Further, for example, the following processing is performed. This is an adjustment of the edge strength.
Ead = Ead1 * Scl
Here, Scl is a scale parameter. Scl may be changed according to luminance Yg or luminance Y. Further, the accuracy of the edge emphasis signal is adjusted so as to match the 8-bit accuracy of the luminance Y after YC conversion on the main image signal side. In this case, an 8-bit signal is used.

The adjusted edge component Ead is added to the luminance Y after YC conversion of the main image signal by the adder 16 to obtain the edge-enhanced luminance Yee.
Yee = Y + Ead
Yee clips to fit in the range [0-255].

  As shown in FIG. 6A, the gamma correction table GTbl inputs a 12-bit signal and outputs a 12-bit signal. As shown in FIG. 6A, the curve has such a curve that the inclination decreases as the luminance before correction increases. On the other hand, the tone control gain table is a table storing the tone control gain according to the input luminance value, and inputs the 12-bit luminance value and outputs the tone control gain. For example, as shown in FIG. 6B, the tone control gain is indicated by a curve that changes in accordance with the luminance value.

  In the configuration of FIG. 5, the Y conversion unit for tone control and the Y conversion unit for edge enhancement processing are one common unit, and therefore the configuration of FIG. 4 is simply added to the configuration of FIG. 1. As a result, the hardware configuration can be reduced, which is advantageous.

  Conventionally, for edge detection, the bit length of the luminance Y is small (8 bits = same as the accuracy of the output image), and the accuracy is too rough. Roughly speaking, edge enhancement is performed by detecting and extracting an edge component present in an image, increasing the edge component by adjustment, and adding it to the original image. When the original image accuracy for edge detection is 8 bits, the detected edge components have the same 8 bit accuracy. If this is increased by adjustment, the accuracy of the edge component becomes coarser than the 8-bit accuracy. When an edge-enhanced image is created by adding a rough edge component to the original image, naturally the smoothness is lost in the edge-enhanced portion.

In this embodiment, since the bit length of the luminance Y used for edge detection is larger than the accuracy of the output image of 8 bits, it is possible to provide sufficient accuracy for edge enhancement.
Further, conventionally, as shown in the gamma correction curve of FIG. 7, the slope of the gamma correction curve is small where the luminance before correction is large. Sometimes gamma correction made it difficult to see.

In the present embodiment, the edge enhancement is performed by adding the adjusted edge component E to the luminance Y after YC conversion, so that the edge enhancement effect is not easily seen.
Conventionally, table data for edge γ gain is required, and adjustment of edge γ is difficult.

In the present embodiment, the table data only needs to be a common gamma correction table for both the main image signal and the edge enhancement signal, and the table data for the edge γ gain is not necessary.
When a change is made to the gamma correction table, in the conventional case, it is necessary to generate table data for the edge γ gain in accordance with the change of the gamma correction table. This is not necessary because a common gamma correction table is used for image signal processing and edge enhancement processing.

In addition to the above embodiment, the following supplementary notes are disclosed.
(Appendix 1)
First gamma correction means for performing gamma correction on the input image signal using a gamma correction table that outputs an output signal of the first accuracy;
Second gamma correction means for performing gamma correction with a second accuracy lower than the first accuracy using the gamma correction table for the input image signal;
Edge extraction means for extracting an edge component of an image signal using an output signal from the first gamma correction means;
Adding means for adding the output of the second gamma correction means and the output of the edge extraction means;
An edge-enhanced image processing apparatus comprising:
(Appendix 2)
The input image signal is an RGB signal,
The second gamma correction means performs gamma correction directly on the input RGB signal,
The edge-enhanced image processing apparatus according to claim 1, further comprising conversion means for converting an output RGB signal from the second gamma correction means into a luminance / color difference signal.
(Appendix 3)
The edge-enhanced image processing apparatus according to appendix 2, wherein the first gamma correction unit performs gamma correction on a luminance signal obtained from an input image signal.
(Appendix 4)
The edge-enhanced image processing apparatus according to appendix 3, wherein the edge extraction unit extracts an edge component from the luminance signal from the first gamma correction unit.
(Appendix 5)
The edge-enhanced image processing apparatus according to appendix 4, wherein the adding means adds the luminance signal from the second gamma correction means and the edge component from the edge extracting means. (Appendix 6)
The edge-enhanced image processing apparatus according to appendix 1, further comprising a gradation correction unit that performs gradation correction of an input image signal before the gamma correction by the first and second gamma correction units.
(Appendix 7)
The second gamma correction means performs gamma correction with the second accuracy, which is less accurate than the first accuracy, by using only the upper bits of the output value of the gamma correction table. The edge-enhanced image processing apparatus according to appendix 1.
(Appendix 8)
The edge-enhanced image processing apparatus according to appendix 1, wherein the edge extraction unit converts the accuracy of the edge component to the second accuracy after extracting the edge component.
(Appendix 9)
The edge-enhanced image processing apparatus according to claim 1, wherein the edge extraction unit includes an edge adjustment unit that adjusts the value of the extracted edge component.
(Appendix 10)
A first gamma correction is performed on the input image signal using a gamma correction table that outputs an output signal of the first accuracy,
Using the gamma correction table for the input image signal, a second gamma correction is performed with a second accuracy lower than the first accuracy,
Using the output signal from the first gamma correction, the edge component of the image signal is extracted,
Adding the second gamma correction result and the edge extraction result;
And an edge-enhanced image processing method.

It is a figure explaining the edge emphasis image processing device of an embodiment of the present invention. It is a figure which shows the conversion graph of the common gamma correction table hold | maintained at the gamma correction parts 10 and 13. FIG. It is a figure explaining a Laplacian filter. It is a figure which shows the structure for tone control (tone correction). It is FIG. (1) explaining 2nd Embodiment containing a tone control. It is FIG. (2) explaining 2nd Embodiment containing a tone control. It is a figure which shows the example of a gamma correction curve. It is a figure explaining the edge emphasis image processing device of the 1st prior art. It is a figure explaining the edge enhancement image processing apparatus of the 2nd prior art. It is a figure explaining the edge emphasis image processing device of the 3rd prior art.

Explanation of symbols

10, 13 Gamma correction unit 11 YC conversion unit 12, 12a Y generation unit 14 Edge detection unit 15 Edge adjustment unit 20 Tone control gain generation unit

Claims (5)

First gamma correction means for performing gamma correction on the input image signal using a gamma correction table that outputs an output signal of the first accuracy;
Second gamma correction means for performing gamma correction with a second accuracy lower than the first accuracy using the gamma correction table for the input image signal;
Edge extraction means for extracting an edge component of an image signal using an output signal from the first gamma correction means;
Adding means for adding the output of the second gamma correction means and the output of the edge extraction means;
An edge-enhanced image processing apparatus comprising: The input image signal is an RGB signal,
The second gamma correction means performs gamma correction directly on the input RGB signal,
2. The edge-enhanced image processing apparatus according to claim 1, further comprising conversion means for converting an output RGB signal from the second gamma correction means into a luminance / color difference signal.   The edge-enhanced image processing apparatus according to claim 2, wherein the first gamma correction unit performs gamma correction on a luminance signal obtained from an input image signal.   The edge-enhanced image processing apparatus according to claim 3, wherein the edge extraction unit extracts an edge component from the luminance signal from the first gamma correction unit.   2. The edge-enhanced image processing apparatus according to claim 1, further comprising a gradation correction unit that performs gradation correction of an input image signal before the gamma correction by the first and second gamma correction units.

Images