KR20090107907A - Method of digital image enhancement based on metadata and image forming appartus - Google Patents

Abstract

PURPOSE: A method of improving a digital image based on metadata and an image forming apparatus thereof are provided to reduce average processing time and not to require additional calculation resources. CONSTITUTION: A method of improving a digital image based on metadata comprises the following steps. Presence probability of JPEG artifacts is determined based on an EXIF-mark. The JPEG artifacts are corrected(S202). Noise presence probability of a high level is determined based on the EXIF-mark. A noise level is measured. The noise level is lowered.

Description

Digital image improvement method and image forming apparatus based on metadata {METHOD OF DIGITAL IMAGE ENHANCEMENT BASED ON METADATA AND IMAGE FORMING APPARTUS}

The claimed invention relates to the field of processing digital images. In particular, the present invention relates to the definition of the quality of an image and its improvement.

With the advent of digital cameras, the processing of photography is becoming more attractive and easier. The number of pictures taken annually increases exponentially, in part due to low cost and simplicity of shooting. Further selection of pictures for storage and printing requires efforts proportional to the number of pictures taken. For example, sorting or storing to select whether it is suitable for printing dozens or even hundreds of photos taken during a trip can be a very difficult task. The quality of a picture is one of the main criteria that a user uses when deciding whether to print a digital picture. Some images may be removed due to insufficient quality (eg, underexposed, overexposed, or low-contrast) or unacceptable content. In general, systems for classifying quality classify an image based on the content-analysis of the image. Each image is generally represented by a low level of a number of attributes (eg, structure, color, and form) obtained from the image. Images are classified while being recorded in qualifiers adjusted for prespecified images according to the type of quality (eg, overexposure, underexposure, low contrast). Based on the attributes entered, the qualifier defines whether a new image should be classified for a given kind of quality.

Photographing made by a digital camera generally includes the camera's metadata, which may include an indication of the sensor, initial data of the camera, and other information about the photographing process. . Metadata is a term related to data concerning data, and for digital photographers using a digital camera, it refers to auxiliary data, which is additional information about the photograph taken. Several years ago, in collaboration with the Association of the Japanese Electronic and Information Industry (JITA), a group of large manufacturers of photographic equipment developed a system of metadata under the name of the Exchange Format of Graphic Files (EXIF). This system was developed to build any set of information about the camera's function, image parameters, and various additional information right within the digital frame at the moment of shooting. This extended set of information can potentially be useful for improving digital photography, but the multidimensional and different structures of the data complicate its use.

Indeed, current digital cameras (Digital Still Cameras (DSC)) can use the additional information area of EXIF, so that metadata is made essential. EXIF's metadata usually serves to store indications of shooting conditions and automatic or manual adjustments. Modern DSC can automatically select the optimal imaging parameters based on the condition of the environment and the set user mode. However, complicated situations of external factors can result in automatic or manual inappropriate selection of the camera. In theory, this can be detected afterwards based on the analysis of the EXIF metadata.

Modern photoprinters can perform printing from a memory card. The user can manually browse the image shown on the DSC or LC display of the printer (image forming apparatus). However, the size of the display described above is generally not sufficient to accurately measure the quality of an image. In addition, many printer models do not have a display at all, and especially considering that the memory card of the current DSC may include hundreds of pictures, the process for selecting the picture images requires a long time.

The printer prints only high-quality photos in full-size, for example, and in the form of low-size sketches for photos that require further processing to improve quality. The sorting operation of the pictures may be performed.

In [3], the authors use a system for the reception of data based on the retrieval of data in order to automatically link the various combinations of multidimensional EXIF metadata with the type of stage. Thus, very simple and effective classification of stages can be performed for any kind of photo. A user interface is also provided so that the user can enter a question about metadata from the keyboard, and the system can find the images corresponding to the question in the base of the graphic data and display it on the display. This approach involves EXIF-meta data with a particular stage type, such as a stage with backlight, an evening stage, and a stage with snow. In order to improve the results of the classification, the initial classification based solely on metadata is combined with a simple analysis based on histograms for rapid identification of detected data. The results of the classification can then be used for best control, estimation, or improvement of the quality of the picture.

US Patent Application 2007-0041657 [4] defines the quality of an image, a separation block that separates the given images into a set of separate images, a calculation block that calculates a histogram of the brightness of the corresponding separated images, a histogram of brightness It will be described with respect to an image processing apparatus including a measuring block for measuring the quality of the separated images using the, and measuring the data quality of the image according to the coefficient oligopoly of the measured quality. Thus, the data quality of the image can be determined more accurately.

U.S. Patent Application Publication 20060239583 [5] allows information on the numerical distribution of pixels of brightness to be determined and corrects the brightness of a specific area of the image using properties according to the characteristics of the distribution of the detected numerical information. A method and apparatus for correcting brightness are described. Thus, deterioration of the features of the recognition of the image by the person is compensated in a specific area, such as a low brightness area of the image.

The profile of the noise of the device represents a reused analysis of the noise properties of an image capture device (digital camera, scanner, etc.) operating in some mode. The use of a noise profile for a display device operating in automatic mode, that is, the use of a device type EXIF information for capturing an image, and also the use of a preliminarily stored or loaded noise profile for a given display device Profiled noise suppression is achieved. The noise analysis of the image and the quality measurement of the image can be performed automatically after the reception of the file of the digital image, using EXIF-meta data.

The technical decision presented in the above-mentioned invention describes a fast method for automatic image selection for printing high quality images from a base of photographic data. High quality images classified based on the EXIF information can be printed in batch mode without further analysis. The main advantage of this method is that the processing time can be reduced compared to the quality analysis method based on the content analysis of the image. .

US Patent Application 20070253699 [6] relates to a classification system based on metadata. Thus, the result of the classification is received based on the histogram consisting of the metadata associated with the image and the brightness value of the image. Thus, the image is classified (referenced) according to the type of stage, and both the result of classification based on metadata and the classification based on histogram are used. The system for processing an image includes a qualifier of a stage type and a qualifier of a check based on metadata. The first qualifier performs initial classification of the image according to the type of stage, and the second qualifier confirms the result of the initial classification according to a method based on the histogram of the brightness value of the image. Corresponding to this method does not allow to classify an image by quality features.

US Pat. No. 6,895,112 [7] describes a red eye detection method based on a red area detection method by using a mark (tag) of an EXIF section indicating whether or not a flash was used in photographing. The digital camera is used for digital recording in which a mark or marks ( "flash (Flash)" mark of the EXIF) that details whether the shot when the flash is used in the camera for each image is removed for each image in this area. The process for detecting red eye described in [7] allows checking the information heading of the image, and if this heading specifies that no flash is used, then no red eye problem exists. It provides the conclusion that it is not without further analysis of the image. This rapid conclusion saves the time required for processing the image and reduces the number of malfunctions of the red eye indicator. Besides. Header may include (in, for example, "flash (Flash)" section of the marks of the EXIF) information as to whether the reflective light from the flash is detected by the camera as well. If no reflective light is detected, the red-eye detection process can conclude that the light from the flash was not sufficient for shooting, and concluded that the light from the flash was not sufficient to display the effect of red-eye. (So you can quickly solve that red-eye is ineffective without requiring additional analysis of the image).

US Patent Application 20050031224 [8] describes a method for detecting red eye defects and correcting these defects. In part, this method is based on the analysis of metadata used for simple detection of red eye defects in digital images. For example, one or more criteria of exception may be used to conclude that no flash was used. Metadata contained in a digital image can be analyzed as an EXIF mark or a simple mark, and extensive post-processing of the image can result in gradation, sharpness and / or color balance of the tone of the image. may be performed based on such information. Another way of using metadata is to find applications in the photo-processing industry when digital images can be subject to post-processing in order to optimize the results of photo printing. For the definition of the features associated with them in the image, it can also be used for correction of overall parameters such as gradation, sharpness, and color balance of the tones of the image.

Many methods, based on analyzing the content of an image and detecting and removing only one type of defects, are described, for example, in US Patents 6 950 473 [9], 5 923 775 [10], 6 987 530 [11] , 728 414 [12]. However, successive application of these methods consumes a lot of time, and their exact order is not clear.

Most of the systems for classification according to the quality feature classify the images based on the content-analysis of the images. Each image is typically represented by a set of low level features (eg, structure, color, and shape) detected in the image. The images are classified based on the corresponding features by the qualifiers learned in the test images, and sorted according to the type of quality (eg, overexposure, underexposure, low contrast, etc.). The qualifier defines, based on the input characteristics, whether a new image should be classified into the type of the corresponding quality.

Unfortunately, classification methods based on the content of the stage require significant computational resources, including the necessary memory resources, which makes them unsuitable for practical application at the hardware level.

Reference

[1] Rob Fergus, Barun Singh, Aaron Hertzmann, Sam T. Roweis, William T. Freeman Removing Camera Shake from a Single Photograph. SIGGRAPH 2006, Boston.

[2] Peter Meer, Jean-Michel Jolion, Azriel Rosenfeld A Fast Parallel Algorighm for Blind Estimation of Noise Variance. IEEE Trans. Pattern Anal. Mach. Intell. 12 (2): 216-223 (1990).

[3] Jonathan Yen, Peng Wu, Daniel Tretter Knowledge discovery for better photographs Proc. of SPIE-IS & T Electronic Imaging, SPIE Vol. 6506, 65060B, 2007.

[4] United States Patent Application Publication No. 20070041657

[5] United States Patent Application Publication 20060239583

[6] United States Patent Application Publication No. 20070253699

[7] United States Patent 6,895,112

[8] United States Patent Application Publication 20050031224

[9] United States Patent 6,950,473

[10] United States Patent 5,923,775

[11] United States Patent 6,987,530

[12] United States Patent 6,728,414

The problem associated with the claimed invention consists in the development of an overall process for classifying standard defects based solely on the use of the EXIF mark, which reduces the average time of processing and does not require additional computational resources.

Technical results are achieved by creating a new method for improving the quality of a digital image based on metadata, the method comprising determining a probability of the presence of JPEG artifacts based on an EXIF-mark, correcting JPEG artifacts Performing a step of determining the probability of the presence of a high level of noise based on the EXIF-mark, measuring the level of the noise, reducing the level of the noise, and exifing the probability of the presence of a blurred motion effect. Determining based on the mark, measuring the level of the blurred motion effect, performing correction of the blurred motion effect, measuring the exposure of the image based on the EXIF-mark of the sketch (thumbnail) of the image, exposure Performing a correction of the color, determining the probability of the presence of red eye on the basis of the EXIF-mark, and And a step of performing a process.

When performing the claimed method, the probability of the presence of the artifacts is JPEG, the EXIF- mark such as "compressed bits per pixel (CompressedBitsPerPixel)", "pixel size Y (PixelYDimension)", "pixel size X (PixelXDimension)" It is preferred to be defined on the basis.

When performing the claimed method, the probability of the presence of the high-level noise, "exposure time (ExposureTime)", "F-number (FNumber)", "ISO speed rating (ISOSpeedRating)", "gain control (GaingControl) Preferably defined based on an EXIF-mark such as "

When performing the claimed method, the probability of the presence of the blurred movement effect is preferably defined based on EXIF- mark such as "exposure time (ExposureTime)".

When carrying out the claimed method, it is desirable to measure the exposure of the image based on an EXIF-mark, such as a sketch (thumbnail) of the image.

When performing the claimed method, is preferably defined based on the probability of the presence of the red-eye effect, the EXIF- mark, such as "flash (Flash)".

On the other hand, the method for improving the quality of the digital image according to another embodiment, the step of receiving EXIF metadata of the digital image, based on the EXIF-mark of the received EXIF metadata defect probability of the digital image Calculating, and performing correction of the defect based on the calculated defect probability.

In this case, the defect is at least one of a JPEG artifact, a high level of noise, a blurry motion effect, an underexposure, an overexposure, a backlight, and a red-eye phenomenon.

On the other hand, the JPEG artifacts, is determined based on the EXIF- mark of "compressed bits per pixel (CompressedBitsPerPixel)", "pixel size Y (PixelYDimension)", and "pixel size X (PixelXDimension)", the high level of the noise, the EXIF- mark of "exposure time (ExposureTime)", is determined on the basis of the "F-number (FNumber)", "ISO speed rating (ISOSpeedRating)", and "gain control (GainControl)", the hazy The movement effect is determined based on the " ExposureTime " in the EXIF-mark, and the underexposure is measured based on the "sketch (thumbnail)" in the EXIF-mark, and the red-eye phenomenon is It is desirable to determine based on " Flash " of the EXIF-marks.

On the other hand, the image forming apparatus according to the embodiment detects a defect by calculating a defect probability of the digital image based on a receiving unit for receiving EXIF metadata of the digital image, and an EXIF-mark of the received EXIF metadata. And a correction unit configured to correct the defect based on the calculated defect probability.

In this case, the defect is preferably at least one of JPEG artifacts, high levels of noise, blurry motion effects, underexposure, overexposure, backlight, and red eye.

On the other hand, the calculation unit, based on the " CompressedBitsPerPixel ", "Pixel Y Dimension (Pixel YDimension )", and "Pixel X Dimension ( PixelXDimension )" of the EXIF-mark , the probability of the presence of JPEG artifacts probability of the presence determination module, wherein the mark of EXIF- JPEG artifacts to determine "exposure time (ExposureTime)", "F-number (FNumber)", "ISO speed rating (ISOSpeedRating)" based on, and "gain control (GainControl)" A noise presence probability determining module for determining a high level of noise presence probability, a presence probability determining module of a blurred motion effect determining a presence probability of a blurred motion effect based on an " ExposureTime " in the EXIF-mark, An exposure measurement module for measuring exposure of an image based on a "sketch (thumbnail)" in the EXIF-mark, and a red-eye to determine the presence probability of red eye based on " Flash " in the EXIF-mark It may include the presence probability determination module of the phase.

In this case, the correction unit, a JPEG artifact correction module for performing the correction of the JPEG artifacts of the digital image, according to the existence probability of the JPEG artifacts determined in the existence probability determination module of the JEPG artifacts, determined in the existence probability determination module of the noise A noise level reduction module for measuring a noise level of the digital image and reducing the noise level of the digital image according to a high level of noise presence probability, and a probability of the blurry motion effect determined in the presence confirmation determination module of the blurry motion effect A blurry motion effect correction module for measuring the level of blurry motion effect of the digital image and correcting the blurry motion effect of the digital image according to the measured exposure of the exposure measurement module. Perform exposure correction And a red-eye correction module configured to perform red-eye correction according to the determined red-eye occurrence probability of the red-eye phenomenon determination module.

For the understanding of the claimed invention, further details thereof are presented together with the graphic materials.

The problem of automatically analyzing and improving the quality of images becomes more realistic with the growing popularity of digital photography. The problem is that the user has a large number of image-taking opportunities, frequently stores data on a computer, and often does not have a realistic opportunity to sort "high quality" images from low quality images. The claimed technical decision represents one of the first steps in this direction.

The first stage for the implementation of the claimed method is the identified selection of the most widely used kind of possible defects. .

The most widely used types of defects can be presented in the following list:

1. Issues of Exposure

Underexpose;

Overexpose;

Low-contrast;

Backlight of the target.

2. << noisy >> picture

3. Color shades (distortion)

Dark blue shades;

Green shade;

Yellow shade;

Red shade.

4. Blurred photo

Defocused picture

Effect of the blurred motion

5. JPEG-artifacts

6. Red eyes

For the development of a classification including the above-mentioned types, a training base of test images was created. The training base included 1054 images. In the table shown in Fig. 1, for each kind of defects, the most likely combinations of mark-features are derived as a result. The table shows the frequency (percentage) of the occurrence of each mark in the training bay of the images.

It is possible to note that self-descriptiveness is not the same for each of the selected EXIF-fields. For example, field cases for "Exposure time (ExposureT im)" and "F-number (FNumber)" The focal length of the effective value of the time of exposure and camera, the mark "brightness value (BrightnessValue)" is brightness It often includes data detailing that the actual value cannot be measured. Thus, the field "brightness value (BrightnessValue)" the image from 8.8%, even more limited in that the mark images are accompanied with useful information. In addition, many marks are replaced by the manufacturers of digital cameras in the Maker Notes field, which makes this information inaccessible in in-line processing.

For efficient classification, it is necessary to select and analyze the most informative combinations of mark-features for each kind of defects. Then, based on such a combination of features, it is necessary to develop the rules of classification and to train the qualifier on the test base of the picture.

Using EXIF information, it is clearly possible to detect the presence or absence of various kinds of defects with a high degree of accuracy. An example of the analysis of the most frequently encountered defects of digital photographs is derived as a result in detail below based on EXIF-information.

Classification of the problem of exposure . As our analysis shows, most of the images that contain the EXIF-marks include sketches (thumbnails), ie, reduced images that are used for designation of the main image. For example, out of 1054 images in our database, 98.5% of the images contain sketches. Analysis of the quality of an image using the histogram << Sketch of Image >>("thumbnailimage") can be guided by the methods described in [4].

<< Noisy >> Analysis based on EXIF -mark for classification of noisy ( noisy picture classification) . In theory, "noise" in the image appears in the case of long exposure, weak illumination of the matrix, high values of ISO, and also excessive amplification of the signal. By preceding it, we, and then the EXIF label, i.e., "exposure time (ExposureTime)", "F-number (FNumber)", "flash (Flash)", "ISO speed rating (ISOSpeedRating,)", "Gain Control ( GainControl ) ".

<<>> noisy one hour exposure of small value, such as the generation of the image will not happen (large values -log2 ( "Exposure Time (ExposureTime)")) sites (site), and of the "F-number smaller value with ( FNumber ) " (the aperture is open). In the specific models of the camera (digital photographing device), even in the whole kind of such device, such a site may be important, but when analyzing all the models as a whole, this site can be determined as follows:

-log2 ( "Exposure Time (ExposureTime)")> 6, and "F-number (FNumber)" <4

( "Exposure Time (ExposureTime)" <0.015, and "F-number (FNumber)" <4)

By applying the treatment of training, there is an opportunity to reveal about 15% of the normal picture that cannot be subject to the procedure for correction of noise, which can generally reduce the processing time.

Unfortunately, it turns out that it is impossible to identify noisy pictures with errors below 30-40%. The information about the flash does not add new data to identify the noisy picture.

The mark "ISO SpeedRatings " is present in 51% of the photos included in the training database.

In theory, high values of ISO are associated with high levels of noise. Nevertheless, for inexpensive and compact cameras, the ISO 100 value can be considered large, while for professional cameras, there is an opportunity for normal image reception even above the ISO 400 value. Accordingly, it is conceivable that there are more noisy pictures at a value of ISO of 800 or more in modern cameras.

The mark "ISO SpeedRatings " is present in 29% of the photos included in the training database. Theoretically, the value "" GainControl "= 2" (amplification of the high gain-signal) corresponds to a higher level of noise. In practice, the mark "High Gain Up" corresponds to a noisy image, but at the same time the noisy image also corresponds to another value of " GainControl ".

Rules are provided for identifying such images that do not require correction of noise:

"Exposure Time (ExposureTime)" <0.015, and "F-number (FNumber)" <4 and

"ISO SpeedRatings "<= 800 (if ISO speed ratings exist) and

" GainControl " = 2 (if " GainControl " exists)

The introduction of this condition based on EXIF-data allows eliminating the stage for filtration of noise on 15-20% of the photographs during processing for improvement of quality, which reduces the overall time of processing.

Classification of pictures with blurred motion . A rule is developed, expressed by the following formula:

(" ExposureTime ">"> 0.0678), blurry image

(" ExposureTime " ≤0.0678), normal

This rule makes the error of revalidation equal to 12.9%. Increasing the number of features does not lead to a substantial reduction in error.

JPEG -Classification of images containing artifacts. First, the EXIF mark " CompressedBitsPerPixel " was analyzed. This mark is present in 87% of the photographs from our database of images. The actual ratio "bit per pixel (BPP)" was calculated as follows:

Actual BPP = (size of file in bits) / (width * height)

The actual BPP depends on the content of the image. It is known that the softer the image, the higher the degree of JPEG compression. This means, for example, that a picture with a defect in a blurred image will have a higher degree of compression.

There is a difficulty in separating a good picture, ie a picture having no artifacts caused by JPEG compression, and a bad picture, ie a picture having such an artifact. In the example considered, it may be the result of a deficiency in the procedure of verification (all bad photos were taken during several photo sessions at the same parameters of illumination and exposure).

It seems appropriate to consider a second kind of errors, namely the problem of minimizing a wrongly perceived defect image: such a qualifier can rule out the presence of a defect, and the qualifier designates the picture as good. . In the process of improvement based on EXIF, such qualifier will inform when the occurrence of artifacts is likely to occur and will not allow to use JPEG improvement of good images. Basically, the application of such a classification may precede the application of a weaker classification.

The method of JPEG compression is based on separating the image into non-overlapping squares of size 8 to 8 pixels and applying discrete consine transformation. This means that with the increase in the scale of the image (actual increase in the resolution when printing), in particular the compression distortion in the printing of the image becomes less visible. From this observation, it can be seen that the size of the image (pixel) can be a good discernible feature for the classification of the image.

Thus, it is provided to use the following EXIF marks for classification of images subject to or not subject to distortion in JPEG compression:

-"Compressed Bits Per Pixel ( CompressedBitsPerPixel )"

-"Pixel Y Dimension ( PixelYDimension )"

-"Pixel X Dimension ( PixelXDimension )"

These marks are widely used by the manufacturers of cameras and are presented in every image of our database.

As a training database, 976 images were used, including 873 good, ie high quality, and 103 bad, ie, low, quality pictures. Linear separation qualifiers have been developed. During testing of 200 images (179 good photos, 21 bad photos), the given qualifier showed the following results:

First Type of Errors (FP) 38

Second kind of errors FN 1

Such results demonstrate that the possibility of excluding good images that do not require JPEG enhancement is quite high.

The rules of decision making to determine whether an image is subject to distortion in JPEG compression are shown as follows:

If C ₁ > C ₂ , there is distortion by JPEG compression, and if C ₁ <C ₂ , the image is sufficiently high quality.

The order of performance of the various stages for identification and correction is also important as they affect each other. In FIG. 2, based on the EXIF-meta data, a block diagram of the procedure for the improvement of the image is presented. Receipt of EXIF metadata of a digital image from a camera, a memory card, a CD-ROM, a CD-RW, a DVD or other device capable of storing information or reading it out is performed (step 201). . The algorithm checks for the presence of perceivable JPEG artifacts (step 202). If the artifact is present, the JPEG artifact is corrected (step 203), for example, as described in [12], and identification of visible noise is performed (step 204). If there is visible noise in the image, the degree (level) of noise is determined (step 205). Many methods of defining the degree of noise are known. For example, it is possible to use the method described in [2]. Reduction of the level of noise is performed (step 206). There are many well known methods for reducing the level of noise. The possibility of the presence of blurry motion is determined (step 207). If there is a chance of detection of blurry motion, identifying a blurry image based on the content of the image (content) is performed (step 208). Correction of the blurry image is performed (step 209). Many methods are known for identifying and correcting blurred motion. One such method is described in [1]. Identification and correction of exposure is performed (step 210). The algorithm checks if the flash has been used (step 211). If so, the identification and correction of the effect of the red-eye phenomenon is performed based on published Russian patent application 2006123847 (step 212). If the flash is not turned on, the operation of the algorithm ends.

In FIG. 3 a block diagram of a module for the processing of exposure is presented, which is used in a procedure for the improvement of an image based on EXIF-meta data. Loading of the sketch of the image from the EXIF-meta data is performed (step 301). A histogram calculation of the brightness over the sketch of the image is performed (step 302). Sketch classification of images with normal, underexposed, overexposed, and backlighting is performed based on the techniques described in published patent applications [4] and [5], and possible recording of the metadata of the classified features is performed. (Step 303). An algorithm is used to determine if the image is underexposed (step 304). If so, extensive correction of contrast is performed (step 306). If not, a algorithm is used to determine if the image is overexposed (step 305). If the image is overexposed, a common correction of contrast is performed (step 306). If the image is not overexposed, a defect in the backlight is identified (step 307). If a defect in the backlight is identified, this defect is corrected (step 308). If this kind of defect is not identified, the operation of the module to handle the exposure ends in the procedure for the improvement of the image based on the EXIF-meta data.

Various modifications for the implementation of the procedures set forth in the claimed invention are possible, but all of them should be considered to be under the protection determined in the scope of the claims.

4 is a block diagram illustrating a configuration of an image forming apparatus 100 according to an embodiment of the present invention. Referring to FIG. 4, the image forming apparatus 100 includes a receiver 110, a calculator 120, and a corrector 130. The image forming apparatus 100 may be a copier, a printer, a facsimile, or a multi function peripheral (MFP) that implements a combination of these functions through a single device.

The receiver 110 may receive EXIF metadata of the digital image from a camera, a memory card, a CD-ROM, a CD-RW, a DVD, or another device capable of storing or reading information.

The calculator 120 may calculate a defect probability of the digital image based on the EXIF-mark of the EXIF metadata received from the receiver 110. In detail, the calculation unit 120 may determine the existence probability determination module 121 of the JPEG artifact, the existence probability determination module 122 of the noise, the existence probability determination module 123 of the blurry motion effect, the exposure measurement module 124, and the red eye. The presence probability determining module 125 of the phenomenon may be included. The defect may be at least one of JPEG artifacts, high level noise, blurry motion effects, underexposure, overexposure, backlight, and red eye.

Specifically, the existence probability determination module 121 of the JPEG artifacts is based on " CompressedBitsPerPixel ", "Pixel Y Size ( PixelYDimension )", and "Pixel XDimension " in the EXIF-mark . The probability of the presence of JPEG artifacts can be determined. And, on determining the presence of noise, the probability module 122 EXIF- mark of "exposure time (ExposureTime)", "F-number (FNumber)", "ISO speed rating (ISOSpeedRating)", and "gain control (GainControl)" High probability of noise presence can be determined based on this. In addition, the presence probability determination module 123 of the blurry motion effect may determine the existence probability of the blurry motion effect based on the " ExposureTime " in the EXIF-mark. In addition, the exposure measurement module 124 may measure the exposure of the image based on “sketch (thumbnail)” in the EXIF-mark. In addition, the presence probability determining module 125 of the red eye may determine the presence probability of the red eye based on “ Flash ” among the EXIF marks. The operation in each module has been described above with respect to each defect, so detailed description thereof is omitted.

The corrector 130 may correct the defect based on the defect probability calculated by the calculator 120. In detail, the correction unit 130 may include a JPEG artifact correction module 131, a noise reduction module 132, a blurry motion effect correction module 133, an exposure correction module 134, and a red eye correction module 135. Can be.

In detail, the JPEG artifact correction module 131 may perform correction of JPEG artifacts of the digital image according to the existence probability of the JPEG artifacts determined by the existence probability determination module 121 of the JEPG artifact. The noise level reduction module 132 may measure the noise level of the digital image and reduce the noise level of the digital image according to the high level noise existence probability determined by the noise probability determining module. The blurry motion effect correction module 133 measures the level of the blurry motion effect of the digital image and measures the blurry motion effect of the digital image according to the presence probability of the blurry motion effect determined by the blurring motion effect determination module 123. I can correct it. The exposure correction module 134 may perform exposure correction of the digital image according to the measured exposure of the exposure measurement module 124. The red eye correction module 135 may perform red eye correction according to the determined red eye occurrence probability of the red eye phenomenon determination module 125. Specific operations of each module have been described above with reference to FIGS. 2 and 3, and thus a detailed description thereof will be omitted.

Accordingly, the image forming apparatus 100 can detect defects of various digital images by using the EXIF-mark of the meta data, thereby reducing the average time of processing.

Although preferred embodiments of the present invention have been illustrated and described above, the present invention is not limited to the specific embodiments described above, and the present invention belongs to the present invention without departing from the gist of the present invention as claimed in the claims. Various modifications can be made by those skilled in the art, and such changes are within the scope of the claims.

1 is a table showing the frequency (percentage) of occurrence of an EXIF-mark and the most likely combination of mark-features for each type of defects in the training database of the image,

2 is a flowchart of a process for improving an image based on an EXIF-mark,

3 is a flow chart of a module for processing exposure for processing for improvement of an image based on the EXIF-mark, and

4 is a block diagram showing the configuration of the image forming apparatus according to the present embodiment.

Claims (14)

A method for improving the quality of a digital image based on metadata, Determining a probability of the presence of a JPEG artifact based on the EXIF-mark; Performing correction of the JPEG artifacts; Determining a high level of noise presence probability based on the EXIF-mark; Measuring the level of noise; Reducing the level of noise; Determining a probability of existence of a blurry motion effect based on the EXIF-mark; Measuring a level of blurry motion effect; Correcting blurry motion effects; Measuring the exposure of the image based on the EXIF-mark of the sketch (thumbnail) of the image; Performing a correction of the exposure; Determining a probability of existence of red eye based on the EXIF-mark; And Red-eye correction is performed. The method of claim 1, The presence of the JPEG artifacts is determined based on EXIF-marks such as " CompressedBitsPerPixel ", "Pixel Y Size ( PixelYDimension )", "Pixel XDimension ". The method of claim 1, Noise presence probability of the high level is determined on the basis of the EXIF- mark such as "exposure time (ExposureTime)", "F-number (FNumber)", "ISO speed rating (ISOSpeedRating)", "gain control (GainControl)" How to be. The method of claim 1, The probability of the presence of the blurry motion effect is determined based on an EXIF-mark such as " ExposureTime ". The method of claim 1, The exposure of the image is measured based on an EXIF-mark, such as a “sketch (thumbnail)” of the image. The method of claim 1, The probability of the presence of red eye is determined based on an EXIF-mark such as " Flash ." In a method for improving the quality of a digital image, Receiving EXIF metadata of the digital image; Calculating a defect probability of the digital image based on the EXIF-mark of the received EXIF metadata; And Performing correction of the defect based on the calculated defect probability. The method of claim 7, wherein The defect is at least one of a JPEG artifact, a high level of noise, a blurry motion effect, an underexposure, an overexposure, a backlight, and a red eye. The method of claim 8, The JPEG artifacts, the EXIF- mark of "compressed bits per pixel (CompressedBitsPerPixel)", "pixel size Y (PixelYDimension)", and, being determined based on a "pixel size X (PixelXDimension)", Noise of the high-level, is determined based on the above EXIF- mark "exposure time (ExposureTime)", "F-number (FNumber)", "ISO speed rating (ISOSpeedRating)", and "gain control (GainControl)" , The blurry motion effect is determined based on an " ExposureTime " in the EXIF-mark, The underexposure is measured based on “sketch (thumbnail)” in the EXIF-mark, The red-eye phenomenon, characterized in that it is determined on the basis of the EXIF- mark of "flash (Flash)". A receiving unit for receiving EXIF metadata of the digital image; And A calculator for calculating a defect probability of the digital image based on the EXIF-mark of the received EXIF metadata; And And a correction unit configured to correct the defect based on the calculated defect probability. The method of claim 10, And wherein the defect is at least one of a JPEG artifact, a high level of noise, a blurry motion effect, an underexposure, an overexposure, a backlight, and a red-eye phenomenon. The method of claim 10, The calculation unit, Presence of JPEG artifacts that determine the probability of the presence of JPEG artifacts based on “ CompressedBitsPerPixel ”, “Pixel Y Size ( PixelYDimension )”, and “Pixel X Size ( PixelXDimension )” of the EXIF-mark . Probability determination module; Based on the above EXIF- mark "exposure time (ExposureTime)", "F-number (FNumber)", "ISO speed rating (ISOSpeedRating)", and "gain control (GainControl)" for determining the presence probability of the noise level high Presence probability determination module of noise; A presence probability determining module of the blurry motion effect that determines the existence probability of the blurry motion effect based on an " ExposureTime " in the EXIF-mark; An exposure measurement module for measuring exposure of an image based on a “sketch (thumbnail)” of the EXIF-marks; And An image forming apparatus characterized in that it comprises a; presence probability of a red-eye determination module for determining a presence probability of the red eye on the basis of the above EXIF- mark "flash (Flash)". The method of claim 12, The correction unit, A JPEG artifact correction module for performing correction of JPEG artifacts of the digital image according to the existence probability of the JPEG artifacts determined in the existence probability determination module of the JEPG artifact; A noise level reduction module for measuring a noise level of the digital image and reducing the noise level of the digital image according to a high level noise existence probability determined by the noise probability determining module of the noise; A blur motion effect correction module for measuring a blur motion effect level of the digital image and correcting the blur motion effect of the digital image according to the presence probability of the blur motion effect determined by the blur motion effect determination module; An exposure correction module for performing exposure correction of the digital image according to the measured exposure of the exposure measurement module; And And a red-eye correction module configured to perform red-eye correction according to the determined red-eye occurrence probability of the red-eye phenomenon determination module. An image forming apparatus using a method for improving the quality of a digital image based on the meta data of any one of claims 1 to 6.

Images