KR101621614B1 - Method and apparatus for enhancing digital image, and apparatus for image processing using the same - Google Patents

Abstract

A digital image enhancement method performed in a computer according to an exemplary embodiment of the present invention includes the steps of converting a color space of an input digital image from RGB to YCbCr to generate a luminance component (Y component), a Cb component, and a Cr component, Generating an illumination component (L component) by filtering the image signal with a smoothing filter having a predetermined cutoff frequency, generating a reflection component map based on a reflection component obtained by dividing a luminance component by an illumination component for each pixel, Normalizing based on the histogram to generate a corrected brightness component, and converting the corrected brightness component, Cb component, and Cr component from YCbCr to RGB to generate a digital enhancement image.

Description

BACKGROUND OF THE INVENTION 1. Field of the Invention [0001] The present invention relates to a digital image enhancement method and apparatus, and an image processing apparatus using the digital image enhancement method and apparatus.

The present invention relates to image processing, and more particularly, to image enhancement processing.

Obtaining a digital image in a backlit environment with illumination towards the camera behind the object of interest or in an environment where sufficient illumination is not ensured on the surface of the object of interest typically results in the surface of the object of interest being generally dark, It is difficult to recognize the shape and information of the surface.

Accordingly, a variety of digital image enhancement techniques have been proposed for visually enhancing the visual information so that the subject and the visual information in the digital image can be distinguished from each other.

BACKGROUND OF THE INVENTION [0002] BACKGROUND OF THE INVENTION [0003] BACKGROUND OF THE INVENTION [0003] BACKGROUND OF THE INVENTION [

According to the Retinex algorithm, brightness is improved by removing illumination components from the original image, assuming that brightness = reflected component × illuminance.

The retinex algorithm shows sufficient improvement in many cases. However, in the case of color image, Retinex algorithm is performed in each of R, G, and B channels, and Retinex performance results for each channel are appropriately combined. Can be derived. At this time, not only the amount of calculation of three channels is small, but also it is not easy to determine the blending ratio, and if the blending ratio is not appropriate, the quality of the final image deteriorates.

The Retinex algorithm uses the logarithm function in the process of removing the illumination component from the input image. By using the R, G, B gradation and the logarithmic function, the Wein's law determines the natural darkness enhancement But the degree of enhancement of the contrast is different between the dark portion and the bright portion, and the dynamic range is also narrowed. In addition, the Lennitex algorithm is based on the assumption that the histogram of the image from which the illumination component is removed follows a normal distribution. However, there is no guarantee that the histogram of the image from which the illumination component is removed will follow the normal distribution. The effect of improving dark areas is insignificant.

Accordingly, there is a need for an image enhancement technique capable of achieving a dark area enhancement effect even when the histogram of the image from which the illumination component is removed does not follow the normal distribution without using a logarithmic function, which is less sensitive to the blending ratio,

SUMMARY OF THE INVENTION It is an object of the present invention to provide an image enhancement apparatus and method for a dark region in a backlight environment.

An object of the present invention is to provide an image processing apparatus using a digital image enhancement method.

SUMMARY OF THE INVENTION It is an object of the present invention to provide a digital image enhancement method and apparatus which have a small amount of computation and are insensitive to a composition ratio, and an image processing apparatus using the same.

SUMMARY OF THE INVENTION It is an object of the present invention to provide a digital image enhancement method and apparatus, and a video processing apparatus using the same, which do not use a log function,

SUMMARY OF THE INVENTION An object of the present invention is to provide an image enhancement apparatus and method capable of obtaining a dark region enhancement effect even when a histogram of an image from which an illumination component is removed does not conform to a normal distribution, and an image processing apparatus using the same.

The solution to the problem of the present invention is not limited to those mentioned above, and other solutions not mentioned can be clearly understood by those skilled in the art from the following description.

According to an aspect of the present invention, there is provided a digital image enhancement method performed in a computer,

Converting a color space of an input digital image from RGB to YCbCr to generate a luminance component (Y component), a Cb component, and a Cr component;

Filtering the luminance component with a smoothing filter having a predetermined cutoff frequency to generate an illumination component (L component);

Generating a reflection component map for each pixel based on a reflection component obtained by dividing the luminance component by the illumination component;

Normalizing based on a histogram of the reflection component map to generate a corrected luminance component; And

And converting the corrected luminance component, the Cb component, and the Cr component from YCbCr to RGB to generate a digital enhanced image.

According to one embodiment, the smoothing filter may be implemented by a sinc function filter, a Butterworth low-pass filter, or a Gaussian low-pass filter.

According to one embodiment, the step of generating the illumination component comprises:

Generating a luminance component of the frequency domain by discrete Fourier transforming the luminance component of the time domain;

Filtering the discrete Fourier transformed frequency domain luminance component with a smoothing filter having a predetermined cutoff frequency in the frequency domain; And

And generating the illumination component by inverse discrete Fourier transforming the luminance component of the filtered frequency domain and inverse-transforming the luminance component of the filtered frequency domain into the time domain.

According to one embodiment, the discrete Fourier transform is a fast Fourier transform and the inverse discrete Fourier transform may be an inverse fast Fourier transform.

According to one embodiment, the step of normalizing based on the histogram of the reflection component map to generate a corrected luminance component comprises:

And trimming and normalizing based on the histogram of the reflection component map to generate a corrected luminance component.

According to one embodiment, the step of normalizing based on the histogram of the reflection component map to generate a corrected luminance component comprises:

Generating a histogram of the reflection component map;

Generating a first normalized histogram by extending a histogram of the reflection component map to a whole gradation range;

Trimming the edge of the primary normalization histogram and extending the trimming histogram to the entire gradation range to generate a second normalized histogram; And

And updating the gray level value of each pixel of the reflection component map according to the second normalization histogram to generate the corrected brightness component.

According to one embodiment, the boundary gradation level of the edge to be trimmed may be determined as the gradation level when the minimum gradation level or the histogram area from the maximum gradation level to the specified gradation level reaches a predetermined area threshold value.

According to one embodiment, the area threshold value of the boundary gradation level for the low edge to be trimmed and the area threshold value of the boundary gradation level for the high edge may be set to be different from each other.

According to one embodiment, the boundary gradation level of the edge to be trimmed may be determined to be a gradation level at which the histogram value of the specific gradation level is equal to the predetermined height threshold value.

According to one embodiment, the step of normalizing based on the histogram of the reflection component map to generate a corrected luminance component comprises:

And generating a corrected luminance component according to a normalized histogram stretched once after trimming the histogram of the reflection component map.

A computer-readable recording medium according to another aspect of the present invention may be a computer-readable recording medium on which a program for implementing a digital image enhancement method according to an embodiment is stored.

According to another aspect of the present invention,

A first conversion unit for converting a color space of an input digital image from RGB to YCbCr to generate a luminance component (Y component), a Cb component, and a Cr component;

A smoothing filtering unit for filtering the luminance component with a smoothing filter having a predetermined cutoff frequency to generate an illumination component (L component);

A reflection component map generator for generating a reflection component map based on a reflection component obtained by dividing the luminance component by the illumination component for each pixel;

A histogram normalizing unit for normalizing the image based on the histogram of the reflection component map to generate a corrected luminance component; And

And a second conversion unit for converting the corrected luminance component, the Cb component, and the Cr component from YCbCr to RGB to generate a digital enhanced image.

According to one embodiment, the smoothing filter may be implemented by a sinc function filter, a Butterworth low-pass filter, or a Gaussian low-pass filter.

According to one embodiment, the smoothing filtering unit comprises:

The luminance component of the time domain is subjected to discrete Fourier transform to generate the luminance component of the frequency domain,

Filtering the frequency components of the discrete Fourier transformed frequency domain with a smoothing filter having a predetermined cutoff frequency in the frequency domain,

And inverse discrete Fourier transforms the luminance component of the filtered frequency domain to inverse time domain to generate the illumination component.

According to one embodiment, the discrete Fourier transform is a fast Fourier transform and the inverse discrete Fourier transform may be an inverse fast Fourier transform.

According to an embodiment, the histogram normalization unit may include:

And trimming and normalizing based on the histogram of the reflection component map to produce a corrected luminance component.

According to an embodiment, the histogram normalization unit may include:

Generates a histogram of the reflection component map,

A first normalized histogram is generated by extending the histogram of the reflection component map in accordance with the entire gradation range,

A second normalized histogram is generated by trimming the edge of the first normalized histogram, stretching the trimming histogram to the entire gradation range,

And to update the gray level values of the pixels of the reflection component map according to the secondary normalization histogram to generate a corrected brightness component.

According to one embodiment, the boundary gradation level of the edge to be trimmed may be determined as the gradation level when the minimum gradation level or the histogram area from the maximum gradation level to the specified gradation level reaches a predetermined area threshold value.

According to one embodiment, the area threshold value of the boundary gradation level for the low edge to be trimmed and the area threshold value of the boundary gradation level for the high edge may be set to be different from each other.

According to one embodiment, the boundary gradation level of the edge to be trimmed may be determined to be a gradation level at which the histogram value of the specific gradation level is equal to the predetermined height threshold value.

According to an embodiment, the histogram normalization unit may include:

And to generate a corrected luminance component according to a normalized histogram stretched once after trimming the histogram of the reflection component map.

According to another aspect of the present invention,

A preprocessing unit for generating a preprocessed digital image;

The color space of the preprocessed digital image is converted from RGB to YCbCr to generate a luminance component, a Cb component, and a Cr component, and the luminance component of the converted YCbCr components is filtered by a smoothing filter having a predetermined cutoff frequency, Generates a reflection component map based on a reflection component obtained by dividing the luminance component by the illumination component for each pixel, normalizes the reflection component map based on the histogram of the reflection component map to generate a corrected luminance component, An image enhancement unit for converting a luminance component, a Cb component, and a Cr component from YCbCr to RGB to generate a digital enhanced image; And

And a main function processing unit for processing a predetermined main function based on the digital enhancement image.

According to one embodiment, the pre-

Processing function of at least one of luminance adjustment, contrast adjustment, sharpness adjustment, tint adjustment, alpha synthesis, gamma correction, artifact or dirt removal, resolution adjustment, distortion correction, background or foreground separation Can operate.

According to one embodiment, the main function processing unit comprises:

At least one of a character recognition function, a license plate recognition function, a road recognition, a sign recognition, an object recognition function, an object tracking function, a pedestrian counting function or a video surveillance function can be performed based on the digital enhanced image.

According to the digital image enhancement method and apparatus and the image processing apparatus using the digital image enhancement method of the present invention, the quality of the digital image can be improved with respect to the dark part of the backlight environment.

According to the digital image enhancement method and apparatus and the image processing apparatus using the digital image enhancement method and apparatus of the present invention, it is possible to improve the quality of a digital image without requiring a low calculation amount and sensitivity to a blending ratio.

According to the digital image enhancement method and apparatus of the present invention and the image processing apparatus using the digital image enhancement method and apparatus, there is no phenomenon that processing of the dark portion and the bright portion is different without using the log function.

According to the digital image enhancement method and apparatus and the image processing apparatus using the digital image enhancement method and apparatus of the present invention, even when the histogram of the image from which the illumination component is removed does not follow the normal distribution, the dark area enhancement effect can be obtained.

The effects of the present invention are not limited to those mentioned above, and other effects not mentioned can be clearly understood by those skilled in the art from the following description.

1 is a flowchart illustrating a digital image enhancement method according to an exemplary embodiment of the present invention.
2 is a flowchart illustrating specific procedures in the smoothing filtering step of the digital image enhancement method according to an exemplary embodiment of the present invention.
3 is a flowchart illustrating specific procedures in the histogram trimming normalization step of the digital image enhancement method according to an exemplary embodiment of the present invention.
FIGS. 4 and 5 are graphs illustrating trimming of a histogram in a histogram trimmed normalization step of the digital image enhancement method according to an exemplary embodiment of the present invention.
FIGS. 6, 7, and 8 are views comparing result images according to the digital image enhancement method according to an embodiment of the present invention with original images and result images processed according to the conventional Retinex technique, respectively.
9 is a block diagram illustrating an image enhancement apparatus according to another embodiment of the present invention.
10 is a block diagram illustrating an image processing apparatus according to another embodiment of the present invention.

For the embodiments of the invention disclosed herein, specific structural and functional descriptions are set forth for the purpose of describing an embodiment of the invention only, and it is to be understood that the embodiments of the invention may be practiced in various forms, The present invention should not be construed as limited to the embodiments described in Figs.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. The same reference numerals are used for the same constituent elements in the drawings and redundant explanations for the same constituent elements are omitted.

1 is a flowchart illustrating a digital image enhancement method according to an exemplary embodiment of the present invention.

The digital image enhancement method of the present invention can be performed in a computer having predetermined information processing capability.

Referring to FIG. 1, in step S11, a digital image enhancement method converts a color space of a digital image from RGB to YCbCr to generate a luminance component (Y component), a Cb component, and a Cr component.

RGB is a color space representing a color as a mixture of red, green and blue components of the three primary colors of light and YCbCr is a color space representing a mixture of a luminance component Y and chrominance components Cb and Cr It is a color space to express.

The relationship of converting the R, G, and B components of a pixel whose R, G, and B components are represented by 8 bits (0 to 255) into Y, Cb, and Cr components is as follows.

Y = krR + kgG + kbB

Cb = B - Y

Cr = R - Y

Cg = G - Y

According to ITU-R BT.601, which is a standard for digital standard (SD) television, kr = 0.299, kg = 0.587 and kb = 0.114, kb + kr + kg = 1. However, ITU-R BT.709, which is a standard for digital high definition (HD) television, has a slightly different value, kr = 0.2126, kg = 0.7152, kb = 0.0722.

Since kg can be expressed in kb and kr, rewriting the above formula can be summarized as follows.

Y = krR + (1 - kb - kr) G + kbB

Cb = 0.5 (B - Y) / (1 - kb)

Cr = 0.5 (R - Y) / (1 - kr)

According to ITU-R BT.601, Y, Cb and Cr can be written as follows.

Y = 0.299R + 0.587G + 0.114B

Cb = 0.5643 (B - Y) + 128

Cr = 0.7132 (R - Y) + 128

Conversely, the relationship between YCbCr and RGB is as follows.

R = Y + (1 - kr) Cr / 0.5

(1 - kb - kr) - 2kr (1 - kr) Cr / (1 - kb - kr)

B = Y + (1 - kb) Cb / 0.5

According to ITU-R BT.601, the following can be written.

R = Y + 1.402 (Cr-128)

G = Y - 0.334 (Cb - 128) - 0.714 (Cr - 128)

B = Y + 1.772 (Cb-128)

In step S12, the luminance component (Y component) of the converted YCbCr components is filtered by a smoothing filter having a predetermined cutoff frequency to generate an illumination component (L component).

The cutoff frequency can be determined empirically and can be determined, for example, in the range of 3 to 10 Hz.

The smoothing filter is a filter for suppressing a frequency component higher than a predetermined cutoff frequency and softening an edge. For example, a sinc filter, a Butterworth low pass filter (BLPF) , And a Gaussian low pass filter (GLPF).

Since the Y component is a luminance component, the very low frequency band of the luminance component (Y component) can be regarded as corresponding to the illumination that is entirely reflected on the subject on the digital image. Therefore, when the luminance component (Y component) is subjected to smoothing filtering with a low cutoff frequency, a luminance component by illumination, that is, an illumination component (L component) can be obtained.

More specifically, referring to FIG. 2 for describing step S12, FIG. 2 is a flowchart illustrating specific procedures in the smoothing filtering step of the digital image enhancement method according to an exemplary embodiment of the present invention.

In Fig. 2, step S12 may be composed of steps S121 to S123.

In step S121, the luminance component (Y component) of the frequency domain is generated by performing Discrete Fourier Transform (DFT) on the luminance component (Y component) of the time domain. Specifically, the discrete Fourier transform (DFT) may be a Fast Fourier Transform (FFT).

In step S122, the luminance component (Y component) of the frequency domain subjected to the discrete Fourier transform is filtered by a smoothing filter having a predetermined cutoff frequency in the frequency domain.

In step S123, the illumination component (L component) is generated by performing inverse discrete Fourier transform (inverse DFT, IDFT) on the luminance component (Y component) of the filtered frequency domain to inverse time domain. Specifically, the inverse discrete Fourier transform (IDFT) may be an inverse fast Fourier transform (IFFT).

Of course, instead of obtaining the illumination component (L component) by performing the smoothing filtering and the inverse Fourier transform again in the frequency domain after the Fourier transform as described above, the luminance component (Y component) and the filtering function are convolution To obtain an illumination component (L component).

Referring back to Fig. 1, in step S13, a reflection component map is generated based on a reflection component obtained by dividing a luminance component (Y component) for each pixel by an illumination component (L component). Each pixel value of the reflection component map may be equal to the reflection component value, or may be a value obtained by calculating a reflection component value by a predetermined arithmetic operation.

In step S14, normalization is performed based on the histogram of the reflection component map to generate a corrected luminance component (Y 'component).

According to the embodiment, in step S14, the corrected luminance component (Y 'component) may be generated by trimming and normalizing based on the histogram of the reflection component map.

3 is a flowchart illustrating exemplary procedures in the histogram trimming normalization step of the digital image enhancement method according to an exemplary embodiment of the present invention. Referring to FIG.

In Fig. 3, step S14 may consist of steps S141 to S143.

In step S141, a histogram of the reflection component map is generated.

In step S142, the histogram of the reflection component map is stretched in accordance with the entire gradation range to generate a first-normalized histogram.

For example, if the entire gradation range is in the range from 0 to 255, and the histogram of the reflection component map generated in step S13 is distributed from the minimum level 10 to the maximum level 100, the histogram value of level 10 The histogram value of the level 100, which was the maximum level, is normalized by stretching proportionally to move to the position of the level 255 at this level 0 position.

In step S143, the edge of the primary normalization histogram is trimmed, and the trimmed histogram is extended to fit the entire gradation range to generate a secondary normalization histogram.

4 and 5, in order to illustrate the operation of step S143, FIGS. 4 and 5 illustrate histogram trimming normalization of the digital image enhancement method according to an exemplary embodiment of the present invention. As shown in FIG.

In FIG. 4, the primary normalized histogram shows a very small histogram value at the low edge or the high edge. The gradation levels of such a very small histogram value are a kind of noise and when the pixels corresponding to the gradation levels of such a small histogram value are distributed among the significant pixels constituting the object of interest, This can be a factor. Thus, eliminating these low frequency gradation levels by trimming both the low edge, high edge, or both edges of this primary normalized histogram can help identify the subject.

According to the embodiment, the border gradation level of the edge to be trimmed may be set such that the histogram area from the minimum gradation level (i.e., gradation level 0) or the maximum gradation level (i.e., gradation level 255) To reach 2% of the entire histogram area.

According to the embodiment, the area threshold value of the boundary gradation level for the low edge and the area threshold value for the boundary gradation level for the high edge can be set to be equal to each other.

According to the embodiment, the area threshold value of the boundary gradation level for the low edge and the area threshold value of the boundary gradation level for the high edge can be set to be different from each other.

Meanwhile, the primary normalization histogram illustrated in FIG. 4 is a normal distribution or has a distribution similar to a normal distribution. In reality, however, it is difficult to expect that the histogram of the reflection map will always have a normal distribution or a similar unimodal distribution. For example, a multimodal distribution with local peaks often at the edges, ).

In FIG. 5, pixels having gradation levels corresponding to local peaks are not only noise, but only a relatively small frequency compared to the middle gradation levels, and may actually be some meaningful pixel.

In this case, the border gradation level of the edge to be trimmed may be determined to be a gradation level such that the histogram value of the specific gradation level corresponds to a predetermined height threshold value, for example, 0.2% of the entire histogram area.

According to the embodiment, the height threshold value of the boundary gradation level for the low edge and the height threshold value of the boundary gradation level for the high edge can be set to be equal to each other.

According to the embodiment, the height threshold value of the boundary gradation level for the low edge and the height threshold value of the boundary gradation level for the high edge can be set to be different from each other.

Returning to Fig. 3, in step S144, the gray level values of the respective pixels of the reflection component map are updated according to the second normalization histogram to generate the corrected brightness component (Y 'component).

On the other hand, according to the embodiment, instead of first extending and trimming the histogram of the reflection component map and then performing second elongation in the steps S142 to S144, after the histogram of the reflection component map is trimmed, A luminance component (Y 'component) corrected according to a normalized histogram may be generated.

1, the corrected luminance component (Y 'component), the Cb component, and the Cr component are converted from YCbCr to RGB to generate a digital enhanced image in step S15.

Reference is made to FIGS. 6 to 8 to illustrate the effect of improving dark areas of digital enhancement images generated through the digital image enhancement method of the present invention.

6, 7, and 8 are diagrams comparing digital enhancement images according to the digital image enhancement method according to an exemplary embodiment of the present invention, respectively, with an original image and a digital enhancement image processed according to a conventional Retinex technique.

6, left side (a) is the original image and (b) in the center is a digital enhancement image obtained according to the conventional Retinex technique. The digital enhancement image in (b) is not improved to such a degree that the dark part of the sign can be read at all, but only the result that the dark part around the dark part is wasted.

On the other hand, (c) on the right side is a digitally enhanced image according to the digital image enhancement method of the present invention, in which the dark part of the sign is dramatically improved, and the characters in the sign, which are not seen at all in the original image and Retinex- It can be seen that it can be identified. Although the inside of the pillar object is brightened and appears to be distorted, if the pillar object also has a character, the character could be identified as well.

However, in terms of human vision, it seems unnatural compared to the conventional Retinex technique. However, in a field that does not require naturalness such as machine vision, character recognition, and object recognition, Points are not a problem.

In Fig. 7, the left side (a) is the original image, and the center (b) is the digital enhanced image obtained by the conventional Retinex technique. The digital enhanced image of FIG. 6 (b) is improved to a certain extent in the center of the large signboard of the wide signboard as compared to FIG. 6 (b), but is improved to the edge. Also, small signs on big signs still do not improve at all.

On the other hand, (c) on the right side is a digitally enhanced image according to the digital image enhancement method of the present invention, in which the dark part of the sign is entirely improved, and a character is displayed on the edge of a large sign, which is not easily seen in the original image and Retinex- It can be seen that it is clearly identifiable and the characters of small signs can be identified. Likewise, the characters on the lower track have improved.

8, the digitally enhanced image according to the digital image enhancement method of the present invention at (c) on the right side has a dark portion of the sign, It can be seen that the characters in the signs have been improved to a certain degree by dramatically improving.

Table 1 shows the gradation standard deviation values calculated respectively in the character areas of the images of Figs. 6, 7 and 8. The higher the standard deviation, the better the discrimination and the recognition rate because the background and the character are well distinguished.

division
Standard Deviation Original image Retinex technique Invention 6 4.48 8.22 49.86 7 13.43 32.57 40.85 8 6.90 12.23 27.03

The standard deviation in the original images of Figs. 6, 7 and 8 is improved to about twice in the images of the Retinex technique, but in the images by the digital image enhancement method of the present invention, the standard deviation is about 3 times to about 10 times . Therefore, it can be expected that the character recognition rate and the object recognition rate will be greatly improved.

9 is a block diagram illustrating an image enhancement apparatus according to another embodiment of the present invention.

9, the image enhancement device 90 includes a first conversion unit 91, a smoothing filtering unit 92, a reflection component map generation unit 93, a histogram normalization unit 94, and a second conversion unit 95 ).

The first conversion unit 91 converts the color space of the digital image from RGB to YCbCr to generate a luminance component (Y component), a Cb component, and a Cr component.

The smoothing filtering unit 92 filters the luminance component (Y component) among the converted YCbCr components with a smoothing filter having a predetermined cutoff frequency to generate an illumination component (L component).

According to an embodiment, the smoothing filter may be implemented by any one of, for example, a sinc function filter, a Butterworth low-pass filter, and a Gaussian low-pass filter.

According to the embodiment, the smoothing filtering unit 92 can obtain an illumination component (L component) by convoluting the luminance component (Y component) and the filtering function in the time domain.

According to the embodiment, the smoothing filtering unit 92 performs discrete Fourier transform of the luminance component (Y component) in the time domain to generate the luminance component (Y component) in the frequency domain, and outputs the luminance component Y (L component) by performing inverse discrete Fourier transform on the luminance component (Y component) of the filtered frequency domain and inverse-transforming the luminance component (Y component) of the filtered frequency domain into the time domain with respect to the frequency domain can do.

Specifically, the discrete Fourier transform may be a fast Fourier transform, and the inverse discrete Fourier transform may be an inverse fast Fourier transform.

The reflection component map generation section 93 generates a reflection component map based on a reflection component obtained by dividing a luminance component (Y component) for each pixel by an illumination component (L component). Each pixel value of the reflection component map may be equal to the reflection component value, or may be a value obtained by calculating a reflection component value by a predetermined arithmetic operation.

The histogram normalization section 94 normalizes the histogram of the reflection component map based on the histogram to generate a corrected luminance component (Y 'component).

According to the embodiment, the histogram normalization section 94 may trimming and normalize based on the histogram of the reflection component map to generate a corrected luminance component (Y 'component).

Specifically, the histogram normalization section 94 generates a histogram of the reflection component map, generates a first normalized histogram by stretching the histogram of the reflection component map according to the entire gradation range, and at least one edge of the first normalized histogram is trimmed The trimmed histogram is extended to the entire gradation range to generate a second normalized histogram. The gradation values of the respective pixels of the reflection component map are updated according to the second normalized histogram, and the corrected luminance component (Y 'component) is updated .

According to the embodiment, the border gradation level of the edge to be trimmed may be set such that the histogram area from the minimum gradation level (i.e., gradation level 0) or the maximum gradation level (i.e., gradation level 255) To reach 2% of the entire histogram area.

According to the embodiment, the area threshold value of the boundary gradation level for the low edge and the area threshold value for the boundary gradation level for the high edge can be set to be equal to each other.

According to the embodiment, the area threshold value of the boundary gradation level for the low edge and the area threshold value of the boundary gradation level for the high edge can be set to be different from each other.

According to the embodiment, the border gradation level of the edge to be trimmed may be determined as a gradation level such that the histogram value of the specific gradation level corresponds to a predetermined height threshold, for example, 0.2% of the entire histogram area.

According to the embodiment, the height threshold value of the boundary gradation level for the low edge and the height threshold value of the boundary gradation level for the high edge can be set to be equal to each other.

According to the embodiment, the height threshold value of the boundary gradation level for the low edge and the height threshold value of the boundary gradation level for the high edge can be set to be different from each other.

On the other hand, according to the embodiment, the histogram normalization section 94 may generate the corrected luminance component (Y 'component) according to the normalized histogram after trimming the histogram of the reflection component map.

The second conversion unit 95 converts the corrected luminance component (Y 'component), the Cb component, and the Cr component from YCbCr to RGB to generate a digital enhanced image.

10 is a block diagram illustrating an image processing apparatus according to another embodiment of the present invention.

Referring to FIG. 10, the image processing apparatus 100 may include a preprocessing unit 110, an image enhancing unit 120, and a main function processing unit 130.

The preprocessing unit 110 may perform various processes on the input digital image such as brightness adjustment, contrast adjustment, sharpness adjustment, tone adjustment, alpha composition, gamma correction, processing functions such as gamma correction, gamma correction, artifact or salt and pepper noise removal, resolution adjustment, distortion correction, background or foreground separation, and the like. And generate a preprocessed digital image.

The image enhancement unit 120 converts the color space of the preprocessed digital image from RGB to YCbCr to generate a luminance component (Y component), a Cb component, and a Cr component.

Then, the image enhancement unit 120 filters the luminance component (Y component) among the converted YCbCr components with a smoothing filter having a predetermined cutoff frequency to generate an illumination component (L component), and calculates a luminance component Y Component) is divided into an illumination component (L component) and a reflection component map.

According to an embodiment, the image enhancement unit 120 may be filtered with a smoothing filter implemented, for example, by a sinc function filter, a Butterworth low-pass filter, or a Gaussian low-pass filter.

According to an embodiment, the image enhancement unit 120 may obtain an illumination component (L component) by performing a convolution operation on a luminance component (Y component) and a filtering function in a time domain.

According to the embodiment, the image enhancement unit 120 performs discrete Fourier transform of the luminance component (Y component) in the time domain to generate a luminance component (Y component) in the frequency domain, and outputs the luminance component Y (L component) by performing inverse discrete Fourier transform on the luminance component (Y component) of the filtered frequency domain and inverse-transforming the luminance component (Y component) of the filtered frequency domain into the time domain with respect to the frequency domain can do.

Specifically, the discrete Fourier transform may be a fast Fourier transform, and the inverse discrete Fourier transform may be an inverse fast Fourier transform.

Next, the image enhancement unit 120 normalizes the image based on the histogram of the reflection component map, generates a corrected luminance component (Y 'component), and outputs the corrected luminance component (Y' component), Cb component, YCbCr to RGB to generate a digital enhancement image.

According to the embodiment, the image enhancement unit 120 may trim and normalize based on the histogram of the reflection component map to generate a corrected luminance component (Y 'component).

Specifically, the image enhancement unit 120 generates a histogram of the reflection component map, generates a first normalized histogram by stretching the histogram of the reflection component map to the entire gradation range, and at least one edge of the first normalized histogram is trimmed The trimmed histogram is extended to the entire gradation range to generate a second normalized histogram. The gradation values of the respective pixels of the reflection component map are updated according to the second normalized histogram, and the corrected luminance component (Y 'component) is updated .

According to the embodiment, the image enhancement unit 120 may generate the corrected brightness component (Y 'component) according to the normalized histogram after trimming the histogram of the reflection component map.

Finally, the image enhancement unit 120 converts the corrected luminance component (Y 'component), Cb component, and Cr component from YCbCr to RGB to generate a digital enhanced image.

The main function processing unit 130 performs predetermined main functions of the image processing apparatus 100 such as a character recognition function, a license plate recognition function, a road recognition, a sign recognition, an object recognition function, an object tracking function , A pedestrian counting function, or a video surveillance function.

It is to be understood that both the foregoing general description and the following detailed description of the present invention are exemplary and explanatory and are intended to provide further explanation of the invention as claimed. It will be understood that variations and specific embodiments which may occur to those skilled in the art are included within the scope of the present invention.

Further, the apparatus according to the present invention can be implemented as a computer-readable code on a computer-readable recording medium. A computer-readable recording medium includes all kinds of recording apparatuses in which data that can be read by a computer system is stored. Examples of the recording medium include ROM, RAM, optical disk, magnetic tape, floppy disk, hard disk, nonvolatile memory and the like. The computer-readable recording medium may also be distributed over a networked computer system so that computer readable code can be stored and executed in a distributed manner.

90 Image Enhancer
91 first conversion section
92 smoothing filtering unit
93 Reflection component map generating unit
94 histogram normalization unit
95 second conversion section
100 image processing device
110 preprocessing section
120 image enhancement unit
130 main function processor

Claims (25)

A digital image enhancement method performed on a computer,
Converting a color space of an input digital image from RGB to YCbCr to generate a luminance component (Y component), a Cb component, and a Cr component;
Filtering the luminance component with a smoothing filter having a predetermined cutoff frequency to generate an illumination component (L component);
Generating a reflection component map for each pixel based on a reflection component obtained by dividing the luminance component by the illumination component;
Normalizing based on a histogram of the reflection component map to generate a corrected luminance component; And
And converting the corrected luminance component, the Cb component, and the Cr component from YCbCr to RGB to generate a digital enhanced image,
Wherein the step of normalizing the luminance component based on the histogram of the reflection component map to generate the corrected luminance component comprises:
Generating a histogram of the reflection component map;
Generating a first normalized histogram by extending a histogram of the reflection component map to a whole gradation range;
Trimming the edge of the primary normalization histogram and extending the trimming histogram to the entire gradation range to generate a second normalized histogram; And
And updating the gray level values of the pixels of the reflection component map according to the second normalization histogram to generate a corrected brightness component,
The boundary gradation level of the edge to be trimmed of the primary normalization histogram is determined as the gradation level when the minimum gradation level or the histogram area from the maximum gradation level to the specified gradation level reaches a predetermined area threshold value,
Wherein an area threshold value of a boundary gradation level with respect to a lower edge of the edge to be trimmed and an area threshold value of a boundary gradation level with respect to a higher edge are set independently of each other. A digital image enhancement method performed on a computer,
Converting a color space of an input digital image from RGB to YCbCr to generate a luminance component (Y component), a Cb component, and a Cr component;
Filtering the luminance component with a smoothing filter having a predetermined cutoff frequency to generate an illumination component (L component);
Generating a reflection component map for each pixel based on a reflection component obtained by dividing the luminance component by the illumination component;
Normalizing based on a histogram of the reflection component map to generate a corrected luminance component; And
And converting the corrected luminance component, the Cb component, and the Cr component from YCbCr to RGB to generate a digital enhanced image,
Wherein the step of normalizing the luminance component based on the histogram of the reflection component map to generate the corrected luminance component comprises:
Generating a histogram of the reflection component map;
Generating a first normalized histogram by extending a histogram of the reflection component map to a whole gradation range;
Trimming the edge of the primary normalization histogram and extending the trimming histogram to the entire gradation range to generate a second normalized histogram; And
And updating the gray level values of the pixels of the reflection component map according to the second normalization histogram to generate a corrected brightness component,
Wherein the boundary gradation level of the edge to be trimmed of the primary normalization histogram is determined such that a histogram value of a specific gradation level is equal to a gradation level equal to a predetermined height threshold value. A digital image enhancement method performed on a computer,
Converting a color space of an input digital image from RGB to YCbCr to generate a luminance component (Y component), a Cb component, and a Cr component;
Filtering the luminance component with a smoothing filter having a predetermined cutoff frequency to generate an illumination component (L component);
Generating a reflection component map for each pixel based on a reflection component obtained by dividing the luminance component by the illumination component;
Trimming and normalizing the reflection component map based on the histogram of the reflection component map to generate a corrected luminance component; And
And converting the corrected luminance component, the Cb component, and the Cr component from YCbCr to RGB to generate a digital enhanced image,
The boundary gradation level of the edge to be trimmed of the histogram of the reflection component map is determined as the gradation level when the minimum gradation level or the histogram area from the maximum gradation level to the specified gradation level reaches a predetermined area threshold value,
Wherein an area threshold value of a boundary gradation level with respect to a lower edge of the edge to be trimmed and an area threshold value of a boundary gradation level with respect to a higher edge are set independently of each other. A digital image enhancement method performed on a computer,
Converting a color space of an input digital image from RGB to YCbCr to generate a luminance component (Y component), a Cb component, and a Cr component;
Filtering the luminance component with a smoothing filter having a predetermined cutoff frequency to generate an illumination component (L component);
Generating a reflection component map for each pixel based on a reflection component obtained by dividing the luminance component by the illumination component;
Trimming and normalizing the reflection component map based on the histogram of the reflection component map to generate a corrected luminance component; And
And converting the corrected luminance component, the Cb component, and the Cr component from YCbCr to RGB to generate a digital enhanced image,
Wherein the boundary gradation level of the edge to be trimmed of the histogram of the reflection component map is determined such that a histogram value of a specific gradation level is equal to a gradation level equal to a predetermined height threshold value. The method of claim 3 or 4, wherein trimming and normalizing based on a histogram of the reflection component map to generate a corrected luminance component further comprises:
And generating a corrected luminance component according to a normalized histogram stretched once after trimming the histogram of the reflection component map. The method of any one of claims 1 to 4, wherein the smoothing filter is implemented using one of a sinc function filter, a Butterworth low-pass filter, and a Gaussian low-pass filter. delete The method of any one of claims 1 to 4, wherein generating the illumination component comprises:
Generating a luminance component of the frequency domain by discrete Fourier transforming the luminance component of the time domain;
Filtering the discrete Fourier transformed frequency domain luminance component with a smoothing filter having a predetermined cutoff frequency in the frequency domain; And
And generating the illumination component by inverse discrete Fourier transforming the luminance component of the filtered frequency domain and inverse transforming the luminance component of the filtered frequency domain into a time domain. delete delete A computer-readable recording medium containing a program for implementing a digital image enhancement method according to any one of claims 1 to 4 in a computer. A first conversion unit for converting a color space of an input digital image from RGB to YCbCr to generate a luminance component (Y component), a Cb component, and a Cr component;
A smoothing filtering unit for filtering the luminance component with a smoothing filter having a predetermined cutoff frequency to generate an illumination component (L component);
A reflection component map generator for generating a reflection component map based on a reflection component obtained by dividing the luminance component by the illumination component for each pixel;
A histogram normalizing unit for normalizing the image based on the histogram of the reflection component map to generate a corrected luminance component; And
And a second conversion unit for converting the corrected luminance component, the Cb component, and the Cr component from YCbCr to RGB to generate a digital enhanced image,
Wherein the histogram normalization unit comprises:
Generates a histogram of the reflection component map,
A first normalized histogram is generated by extending the histogram of the reflection component map in accordance with the entire gradation range,
A second normalized histogram is generated by trimming the edge of the first normalized histogram, stretching the trimming histogram to the entire gradation range,
And to update the gray level values of each pixel of the reflection component map according to the secondary normalization histogram to generate a corrected brightness component,
The boundary gradation level of the edge to be trimmed of the primary normalization histogram is determined as the gradation level when the minimum gradation level or the histogram area from the maximum gradation level to the specified gradation level reaches a predetermined area threshold value,
Wherein an area threshold value of a boundary gradation level with respect to a lower edge of the edge to be trimmed and an area threshold value with respect to a boundary gradation level with respect to a higher edge are set independently of each other. A first conversion unit for converting a color space of an input digital image from RGB to YCbCr to generate a luminance component (Y component), a Cb component, and a Cr component;
A smoothing filtering unit for filtering the luminance component with a smoothing filter having a predetermined cutoff frequency to generate an illumination component (L component);
A reflection component map generator for generating a reflection component map based on a reflection component obtained by dividing the luminance component by the illumination component for each pixel;
A histogram normalizing unit for normalizing the image based on the histogram of the reflection component map to generate a corrected luminance component; And
And a second conversion unit for converting the corrected luminance component, the Cb component, and the Cr component from YCbCr to RGB to generate a digital enhanced image,
Wherein the histogram normalization unit comprises:
Generates a histogram of the reflection component map,
A first normalized histogram is generated by extending the histogram of the reflection component map in accordance with the entire gradation range,
A second normalized histogram is generated by trimming the edge of the first normalized histogram, stretching the trimming histogram to the entire gradation range,
And to update the gray level values of each pixel of the reflection component map according to the secondary normalization histogram to generate a corrected brightness component,
Wherein a boundary gradation level of the edge to be trimmed of the primary normalization histogram is determined such that a histogram value of a specific gradation level is a gradation level equal to a predetermined height threshold value. A first conversion unit for converting a color space of an input digital image from RGB to YCbCr to generate a luminance component (Y component), a Cb component, and a Cr component;
A smoothing filtering unit for filtering the luminance component with a smoothing filter having a predetermined cutoff frequency to generate an illumination component (L component);
A reflection component map generator for generating a reflection component map based on a reflection component obtained by dividing the luminance component by the illumination component for each pixel;
A histogram normalization unit for trimming and normalizing the reflection component map based on the histogram of the reflection component map to generate a corrected luminance component; And
And a second conversion unit for converting the corrected luminance component, the Cb component, and the Cr component from YCbCr to RGB to generate a digital enhanced image,
The boundary gradation level of the edge to be trimmed of the histogram of the reflection component map is determined as the gradation level when the minimum gradation level or the histogram area from the maximum gradation level to the specified gradation level reaches a predetermined area threshold value,
Wherein an area threshold value of a boundary gradation level with respect to a lower edge of the edge to be trimmed and an area threshold value with respect to a boundary gradation level with respect to a higher edge are set independently of each other. A first conversion unit for converting a color space of an input digital image from RGB to YCbCr to generate a luminance component (Y component), a Cb component, and a Cr component;
A smoothing filtering unit for filtering the luminance component with a smoothing filter having a predetermined cutoff frequency to generate an illumination component (L component);
A reflection component map generator for generating a reflection component map based on a reflection component obtained by dividing the luminance component by the illumination component for each pixel;
A histogram normalization unit for trimming and normalizing the reflection component map based on the histogram of the reflection component map to generate a corrected luminance component; And
And a second conversion unit for converting the corrected luminance component, the Cb component, and the Cr component from YCbCr to RGB to generate a digital enhanced image,
Wherein the boundary gradation level of the edge to be trimmed of the histogram of the reflection component map is determined such that a histogram value of a specific gradation level is equal to a gradation level equal to a predetermined height threshold value. The histogram normalizing unit according to claim 14 or 15,
And to generate a corrected luminance component according to a normalized histogram stretched once after trimming the histogram of the reflection component map. The digital image enhancement apparatus according to any one of claims 12 to 15, wherein the smoothing filter is implemented by any one of a sinc function filter, a Butterworth low-pass filter, and a Gaussian low-pass filter. The apparatus of any of claims 12 to 15, wherein the smoothing filtering unit
The luminance component of the time domain is subjected to discrete Fourier transform to generate the luminance component of the frequency domain,
Filtering the frequency components of the discrete Fourier transformed frequency domain with a smoothing filter having a predetermined cutoff frequency in the frequency domain,
And inverse discrete Fourier transforms the luminance component of the filtered frequency domain and inverse transforms the luminance component of the filtered frequency domain into a time domain to generate the illumination component. 19. The digital image enhancement apparatus of claim 18, wherein the discrete Fourier transform is a fast Fourier transform and the inverse discrete Fourier transform is an inverse fast Fourier transform. delete delete A preprocessing unit for generating a preprocessed digital image;
The color space of the preprocessed digital image is converted from RGB to YCbCr to generate a luminance component, a Cb component, and a Cr component, and the luminance component of the converted YCbCr components is filtered by a smoothing filter having a predetermined cutoff frequency, Generates a reflection component map based on a reflection component obtained by dividing the luminance component by the illumination component for each pixel, normalizes the reflection component map based on the histogram of the reflection component map to generate a corrected luminance component, An image enhancement unit for converting a luminance component, a Cb component, and a Cr component from YCbCr to RGB to generate a digital enhanced image; And
And a main function processing unit for processing a predetermined main function based on the digital enhanced image,
The image enhancement unit
The image processing apparatus according to any one of claims 12 to 15, comprising a digital image enhancement device. 24. The apparatus of claim 22, wherein the pre-
Processing function of at least one of luminance adjustment, contrast adjustment, sharpness adjustment, tint adjustment, alpha synthesis, gamma correction, artifact or dirt removal, resolution adjustment, distortion correction, background or foreground separation The image processing apparatus comprising: 24. The apparatus of claim 22,
Wherein at least one of a character recognition function, a plate recognition function, a road recognition, a sign recognition, an object recognition function, an object tracking function, a pedestrian counting function, or an image surveillance function is performed based on the digital enhanced image.


The digital image enhancement method of claim 8, wherein the discrete Fourier transform is a fast Fourier transform and the inverse discrete Fourier transform is an inverse fast Fourier transform.

Images