JP2006277688A - Apparatus and method for improving image quality, and program therefor - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an apparatus and a method for improving image quality, to automatically adjust luminance according to the luminance information of an input image so as to enable an image output of stable image quality, and also to provide a computer program for executing the method. <P>SOLUTION: The apparatus includes: a bandpass filter section 21 for eliminating noise from the input image, and enhancing contrast using a change in the luminance between neighboring pixels, etc.; an average value correction section 22 for obtaining an average luminance value and correcting the luminance value to satisfy a predetermined condition; and a variance correction section 23 for obtaining a luminance value distribution and correcting the luminance value distribution so as to satisfy a predetermined condition. Thus, an image (of stable image quality) having a luminance value and a dispersion value of the luminance value constantly included within a certain range is generated. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to image quality improvement of captured images, and particularly to an image quality improvement apparatus and image quality that are favorable for an object detection / recognition system using an image obtained by imaging a subject outdoors or in a car where there are fluctuations in the brightness environment The present invention relates to an improvement method and a computer program for executing the method.

  Conventionally, research on object detection / recognition systems using images has been conducted. Hereinafter, in order to make the description easy to understand, the detection / recognition target object is described as a person. Such research aims at improving the performance of security systems and preventing drivers from falling asleep and looking aside by detecting and recognizing human faces.

  In these applications, the person detection / recognition apparatus is required to satisfy certain conditions in the luminance information of the person shown in the image. However, fluctuations in brightness and darkness are severe in the environment such as outdoors or in a vehicle interior, and it is indispensable to generate images with stable image quality for outdoor security devices and driver monitoring devices. is necessary.

As a conventional technique for generating an image with stable image quality, first, a face region is extracted, and the rotation around the camera optical axis is corrected, and then the luminance unevenness on the face surface is corrected. Such a technique is disclosed in Patent Document 1 below.
JP 2001-331795 A (paragraphs 0014 to 0029)

  However, if the brightness value on the face surface is too bright or too dark due to changes in the light and dark environment, it may be difficult to detect the face, and thus the technique disclosed in Patent Document 1 will be difficult to use. Therefore, it is necessary to always generate an image with stable image quality with respect to an image signal input from a camera as a photographing unit.

  SUMMARY OF THE INVENTION The present invention solves the above-described problem, and performs an image quality improvement apparatus, an image quality improvement method, and a method thereof that can always generate an image with stable image quality from an image signal input from a camera. The purpose is to provide a program.

  In order to solve the above-described problems, according to the present invention, the luminance of each main pixel in a plurality of pixels constituting a raster is compared with the luminance of pixels around the pixel in the input image signal. A pixel value calculation unit that obtains a pixel value, an average value correction unit that corrects the pixel value of the image to the target average luminance value, and a variance value correction unit that corrects the pixel value of the image to the target luminance variance value. An image quality improving apparatus, an image quality improving method for performing similar processing, and a program for executing the method are provided.

That is, according to the present invention, the step of obtaining the pixel value by comparing the luminance of each main pixel with the luminance of the pixels around the pixel among the plurality of pixels constituting the raster obtained from the input image signal; ,
Calculating an average value of the pixel values;
A step of correcting the pixel value by shifting the distribution so that the average value in the distribution of the pixel value determined by the pixel value and its frequency matches the predetermined value when the average value deviates from the predetermined value; ,
Obtaining a predetermined range of pixel values as a conversion target pixel value from the distribution of pixel values obtained in the correcting step;
Correcting so that pixels within the range of the conversion target pixel value have a distribution having a predetermined dispersion value,
A method for improving image quality is provided.

According to the present invention, the step of obtaining the pixel value by comparing the luminance of each of the main pixels among the plurality of pixels constituting the raster obtained from the input image signal with the luminance of pixels around the pixel; ,
Calculating an average value of the pixel values;
A step of correcting the pixel value by shifting the distribution so that the average value in the distribution of the pixel value determined by the pixel value and its frequency matches the predetermined value when the average value deviates from the predetermined value; ,
Obtaining a predetermined range of pixel values as a conversion target pixel value from the distribution of pixel values obtained in the correcting step;
Correcting so that pixels within the range of the conversion target pixel value have a distribution having a predetermined dispersion value,
A program for causing a computer to execute the image quality improving method is provided.

Further, according to the present invention, a pixel value for obtaining a pixel value by comparing the luminance of each main pixel with the luminance of pixels around the pixel among a plurality of pixels constituting a raster obtained from an input image signal A calculation unit;
An average value calculation unit for calculating an average value of the pixel values;
When the average value deviates from the predetermined value, a pixel value for correcting the pixel value by shifting the distribution so that the average value in the distribution of the pixel value determined by the pixel value and its frequency matches the predetermined value The correction part of
A conversion target pixel value range calculation unit that obtains a range of pixel values to be converted by deleting pixel values that are equal to or larger than a predetermined value and those that are equal to or smaller than a predetermined value from the distribution of pixel values obtained by the correction unit; ,
A pixel value correction unit that corrects the pixels within the range of the conversion target pixel value to have a distribution having a predetermined dispersion value;
An apparatus for improving image quality is provided.
Note that it is a preferable aspect of the present invention to use a band-pass filter unit that removes noise on the input image and simultaneously improves the contrast of the image as the pixel value calculation unit.

  The image quality improvement apparatus, method, and program of the present invention have the above-described configuration, correct luminance for an image signal input from a camera, and always generate an image with stable image quality even when there is a change in light and dark environments. Make it possible.

  Hereinafter, an image quality improvement apparatus and a processing method according to an embodiment of the present invention will be described with reference to FIGS. FIG. 1 is a diagram showing a schematic configuration of an image quality improvement apparatus as an embodiment of the present invention. 2A and 2B are diagrams showing distributions of pixel values before and after correction by the average value correction unit in the image quality improvement apparatus according to the embodiment of the present invention. FIG. 3 is a diagram showing the configuration of the average value correction unit in the image quality improvement apparatus according to the embodiment of the present invention. FIG. 4 is a diagram showing the configuration of the dispersion value correction unit in the image quality improvement apparatus according to the embodiment of the present invention. FIG. 5 is a diagram illustrating a configuration of a conversion target pixel value range calculation unit in the image quality improvement apparatus according to the embodiment of the present invention. FIG. 6 is a diagram showing a cumulative probability distribution of pixel values in the image quality improvement apparatus according to the embodiment of the present invention. FIG. 7 is a diagram showing an outline of sample value calculation of pixel values in the image quality improvement apparatus according to the embodiment of the present invention. FIG. 8 is a flowchart for explaining an image quality improvement flow in the image quality improvement apparatus according to the embodiment of the present invention.

  First, a schematic configuration of an image quality improvement apparatus according to an embodiment of the present invention will be described with reference to FIG. As shown in FIG. 1, the image quality improvement apparatus 20 includes a band pass filter unit 21, an average value correction unit 22, and a dispersion value correction unit 23. The image input unit 10 takes in a signal indicating an image taken by a camera or the like. The input image signal captured by the image input unit 10 is sent to the band pass filter unit 21. Here, the range of the luminance value of the captured input image signal is 0 to 255. In the present embodiment, the pixel value is represented by an 8-bit digital signal, but the range of the luminance value can be changed according to the number of bits of the digital signal representing the pixel value. A plurality of blocks in the image quality improvement apparatus 20 shown in FIG. 1 and blocks shown in other block diagrams to be described later may be configured as discrete circuits, but a CPU (Central Processing Unit), a memory, and an interface It is realizable with the computer which consists of.

  The bandpass filter unit 21 performs noise removal and enhancement of luminance characteristics (contrast) on the input image sent from the image input unit 10 using a bandpass filter. Here, a Laplacian of Gaussian (hereinafter, also simply referred to as LoG) filter is used as the bandpass filter. Since the LoG filter is a known technique, a detailed description thereof will be omitted. The processing method of the LoG filter is shown in the following formula (1). The performance of the LoG filter depends on the value of σ (a numerical value representing the degree of emphasizing the luminance change with surrounding pixels). As an example, σ = 1 here. The coefficient of the filter mask LoG (i, j) generated when σ = 1 is shown in the following equation (2). This mask size is 7 × 7 pixels. Note that the value of the processed pixel is a real value, and when Expression (2) is used, the range of the processed pixel value is −4590 to +4335. As described above, the band pass filter unit 21 uses the luminance of each main pixel among the plurality of pixels constituting the raster (one screen) obtained from the input image signal as the luminance of the pixels around the pixel. It functions as a pixel value calculation unit that obtains pixel values by comparison. Note that the following equation (2) cannot be applied to pixels at positions near the upper, lower, left, and right outer peripheral portions of the raster, and therefore these pixels are not processed. The processed image is sent to the average value correction unit 22.

The average value correction unit 22 corrects the average value of the pixel values processed by the bandpass filter as shown in FIG. 2A. As illustrated in FIG. 3, the average value correction unit 22 includes an average value calculation unit 222, a correction value calculation unit 222, and a pixel value correction unit 223. First, the average value of the processed pixel values is obtained using the following equation (3). _{Here, Mean_org, I real (n,} m), X, Y is the average value of each uncorrected pixel value, a pixel in the horizontal axis n and the vertical axis m, a width and height of the image. Next, a correction value is calculated by the correction value calculation unit 222 using Mean_org calculated by Expression (3). An example of correction value calculation is shown in the following equation (4). Here, SHIFT and T_Average are a correction value and a target average value to be corrected, respectively. T_Average is a predetermined parameter. Finally, the pixel value is corrected by the pixel value correction unit 223 using the calculated correction value. An example of a pixel value correction method is shown in Expression (5). Here, I _{real_restored} (x, y) and I _real (x, y) are the corrected value and the uncorrected value of the pixels on the horizontal axis x and the vertical axis y, respectively. FIG. 2B shows an example of pixel value distribution after correction. However, in this example, the T_Average value was set to 0. The image whose pixel average value is corrected is sent to the dispersion value correction unit 23.

The dispersion value correcting unit 23 converts and corrects the pixel value so that the luminance dispersion is within a certain range in order to obtain an image with stable image quality. As shown in FIG. 4, the variance value correction unit 23 includes a conversion target pixel value range calculation unit 231, a sample value calculation unit 232, and a pixel value conversion unit 233. Further, as illustrated in FIG. 5, the conversion target pixel value range calculation unit 231 includes a pixel cumulative probability distribution calculation unit 2311 and a range calculation unit 2312. The pixel value (luminance) distribution of the image differs depending on the illumination conditions, the background / object pattern, and the like. For this reason, the range of pixel values to be converted is different for each image. In order to calculate the range of pixel values to be converted (corrected) in each image, the conversion target pixel value range calculation unit 231 first accumulates pixel values as shown in FIG. Find the probability distribution. The cumulative probability distribution of pixel values is obtained by the following equation (6). Here, F _(I real_restored) is the cumulative probability of I _{Real_restored} (Unit: [%]) in, P _(I real_restored) emergence Percentage of I _{Real_restored} the pixel value (Unit: [%]) is. Next, the range calculation unit 2312 obtains a range of pixel values to be converted using the cumulative probability distribution of the calculated pixel values. The calculation method of the conversion target pixel value range is shown in the following formula (7). Here, Intensity_min, Intensity_max, and A are cumulative amounts of the minimum pixel value to be converted, the maximum pixel value to be converted, and the pixel value not to be converted, respectively. A is a predetermined parameter. These information and image are sent to the sample value calculation unit 232.

Next, the sample value calculation unit 232 calculates the number of samples used to convert the pixel value (I _{real_restored} (x, y)), which is a real value, into a discrete value of luminance from 0 to 255. Here, as shown in FIG. 7, the width of the pixel value smaller than the average value (L _Minus ) and the width of the pixel value larger than the average value (L _Plus ) are often not the same width. In order not to deviate from the target luminance value, it is necessary to separately obtain a sample value for pixels having a value smaller than the average value and a sample value for pixels having a value larger than the average value. These sample values are obtained by the following equation (8). Here, Sampling_Minus and Sampling_Plus are a sample value for a pixel having a value smaller than the average value and a sample value for a pixel having a value larger than the average value, respectively (see FIG. 7). B and C are the number of samples for pixels having a value smaller than the average value and the number of samples for pixels having a value larger than the average value, respectively, and are predetermined parameters. This value determines the spread of the luminance distribution of the corrected image. B and C may be equal.

The pixel value conversion unit 233 converts the pixel value using the Intensity_min and Intensity_max calculated by the conversion target pixel value range calculation unit 231 and the Sampling_Minus and Sampling_Plus calculated by the sample value calculation unit 232. The conversion is performed by the following equation (9). Here, I _out (x, y) is a luminance value after conversion, and the range of I _out (x, y) is 0 to 255. D is the luminance center value of the corrected image, and is a predetermined parameter. Round [] means rounding off the first decimal place.

  Next, an image quality improvement flow in the image quality improvement apparatus of the present invention will be described with reference to FIG. The image input unit 10 captures an object or inputs an image from a device (not shown), and the input image is sent to the bandpass filter unit 10 (step S30). The bandpass filter unit 21 performs processing by applying a bandpass filter (here, LoG filter) to the input image (step S31). The average value correction unit 22 calculates an average of pixel values for an image processed by the LoG filter, calculates a deviation of the pixel value from a certain pixel value (target value), and corrects the deviation ( Step S32). The variance value correction unit 23 obtains a range of pixel values to be converted (corrected) for the pixel value whose average value has been corrected, and the pixels within the range have a distribution having a certain variance value (target value). Correction (conversion) is performed as described above (step S33).

  As described above, the image quality improvement apparatus, the image quality improvement method, and the computer program for executing the method according to the present invention can improve the image quality even for an image with a sharp change in luminance. This is useful in the field of various devices that recognize a subject using captured images.

It is a figure which shows the structure of the image quality improvement apparatus which concerns on embodiment of this invention. It is a figure which shows distribution of the pixel value before correction | amendment in the average value correction | amendment part in the image quality improvement apparatus which concerns on embodiment of this invention. It is a figure which shows distribution of the pixel value after correction | amendment in the average value correction | amendment part in the image quality improvement apparatus which concerns on embodiment of this invention. It is a figure which shows the structure of the average value correction | amendment part in the image quality improvement apparatus which concerns on embodiment of this invention. It is a figure which shows the structure of the dispersion value correction | amendment part in the image quality improvement apparatus which concerns on embodiment of this invention. It is a figure which shows the structure of the calculation part of the conversion target pixel value range in the image quality improvement apparatus which concerns on embodiment of this invention. It is a figure which shows the cumulative probability distribution of the pixel value in the image quality improvement apparatus which concerns on embodiment of this invention. It is a figure which shows the sample value calculation outline | summary of the pixel value in the image quality improvement apparatus which concerns on embodiment of this invention. It is a flowchart for demonstrating the image quality improvement flow in the image quality improvement apparatus which concerns on embodiment of this invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 Image input part 20 Image quality improvement apparatus 21 Band pass filter part 22 Average value correction part 23 Dispersion value correction part 221 Average value calculation part 222 Correction value calculation part 223 Pixel value correction part 231 Conversion object pixel value range calculation part 232 Specimen Value calculation unit 233 Pixel value conversion unit 2311 Pixel cumulative probability distribution calculation unit 2312 Range calculation unit

Claims (12)

A step of obtaining a pixel value by comparing the luminance of each of the main pixels among the plurality of pixels constituting the raster obtained from the input image signal with the luminance of pixels around the pixel;
Calculating an average value of the pixel values;
A step of correcting the pixel value by shifting the distribution so that the average value in the distribution of the pixel value determined by the pixel value and its frequency matches the predetermined value when the average value deviates from the predetermined value; ,
Obtaining a predetermined range of pixel values as a conversion target pixel value from the distribution of pixel values obtained in the correcting step;
Correcting so that pixels within the range of the conversion target pixel value have a distribution having a predetermined dispersion value,
A method for improving image quality.   The step of obtaining the conversion target pixel value converts the pixel value distribution obtained in the correcting step into a pixel value equal to or higher than a first predetermined value to the first predetermined value, and the pixel value is a second value The image quality improvement method according to claim 1, wherein the image quality improvement method is executed by converting a value equal to or less than a predetermined value to the second predetermined value.   The image quality improvement method according to claim 1, wherein the step of obtaining the conversion target pixel value includes a step of obtaining a cumulative probability distribution of pixel values.   The step of correcting to a distribution having the predetermined variance value divides the difference between the average value of the pixel values and the minimum value in the range of the pixel value to be converted by a predetermined value in the frequency distribution of pixel values. 4. The image quality improvement according to claim 3, further comprising: obtaining a first sample value, and dividing the difference between the maximum value in the range of the pixel value to be converted and the average value by the predetermined value to obtain a second sample value. Method.   5. The step of correcting the distribution to have a predetermined variance value includes a step of converting a pixel value by a predetermined algorithm using the first sample value and the second sample value. Image quality improvement method. A step of obtaining a pixel value by comparing the luminance of each of the main pixels among the plurality of pixels constituting the raster obtained from the input image signal with the luminance of pixels around the pixel;
Calculating an average value of the pixel values;
A step of correcting the pixel value by shifting the distribution so that the average value in the distribution of the pixel value determined by the pixel value and its frequency matches the predetermined value when the average value deviates from the predetermined value; ,
Obtaining a predetermined range of pixel values as a conversion target pixel value from the distribution of pixel values obtained in the correcting step;
Correcting so that pixels within the range of the conversion target pixel value have a distribution having a predetermined dispersion value,
A program for causing a computer to execute an image quality improvement method.   In the step of obtaining the conversion target pixel value, the pixel value distribution obtained in the correcting step is converted from the pixel value distribution equal to or higher than the first predetermined value to the first predetermined value, and the pixel value is the second value. The program according to claim 6, which is executed by converting a value equal to or less than a predetermined value to the second predetermined value.   The program according to claim 6 or 7, wherein the step of obtaining the conversion target pixel value includes a step of obtaining a cumulative probability distribution of pixel values.   The step of correcting to a distribution having the predetermined variance value divides the difference between the average value of the pixel values and the minimum value in the range of the pixel value to be converted by a predetermined value in the frequency distribution of pixel values. The program according to claim 8, further comprising: obtaining a first sample value, and dividing a difference between a maximum value in the range of the conversion target pixel value and the average value by the predetermined value to obtain a second sample value.   The step of correcting the distribution so as to have a predetermined variance value includes a step of converting a pixel value by a predetermined algorithm using the first sample value and the second sample value. Program. A pixel value calculation unit that obtains a pixel value by comparing the luminance of each main pixel with the luminance of pixels around the pixel among a plurality of pixels constituting a raster obtained from an input image signal;
An average value calculation unit for calculating an average value of the pixel values;
When the average value deviates from the predetermined value, a pixel value for correcting the pixel value by shifting the distribution so that the average value in the distribution of the pixel value determined by the pixel value and its frequency matches the predetermined value The correction part of
A conversion target pixel value range calculation unit that obtains a range of pixel values to be converted by deleting pixel values that are equal to or larger than a predetermined value and those that are equal to or smaller than a predetermined value from the distribution of pixel values obtained by the correction unit; ,
A pixel value correction unit that corrects the pixels within the range of the conversion target pixel value to have a distribution having a predetermined dispersion value;
A device for improving image quality. 12. The image quality improving apparatus according to claim 11, wherein the pixel value calculating unit uses band-pass filter means for removing noise on an input image and simultaneously improving image contrast.

Images