JP2004061500A - Method of detecting image defect - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To precisely detect an image defect caused by a foreign matter deposited on a film, a flaw in the film, a foreign matter deposited on a reading system for reading the film or a photographing system for photographing a subject, or the like, without making erroneous detection. <P>SOLUTION: An image defect candidate, of an image reproduced based on an image data, caused by a damage flaw of a document and the foreign matter deposited on the document, and the foreign matter deposited on the reading system for reading the document or the photographing system for photographing the subject, is detected out of the image data provided by the photoelectric reading for the document or the photoelectric photographing for the subject, the image defect candidate existing in a place where density exceeds a prescribed condition is removed in response to the density where the image defect candidate exists, and the rest is determined as the image defect to solve the problem herein. <P>COPYRIGHT: (C)2004,JPO

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention belongs to the technical field of digital image analysis, and more specifically, an image obtained by photoelectrically reading an image photographed on a film or the like, or a subject is photographed photoelectrically by a digital still camera (DSC). Reading system that reads foreign matter such as dust attached to the film from the image, film scratches and film, especially foreign matter such as dust attached to the image sensor such as CCD, or photographing system for photographing the subject, especially image sensor such as CCD The present invention relates to a method for detecting an image defect that can accurately detect an image defect caused by foreign matter such as dust adhering to the surface.
[0002]
[Prior art]
At present, printing of an image taken on a photographic film (hereinafter referred to as a film) such as a negative film or a reversal film onto a photosensitive material (photographic paper) is performed by projecting an image on the film onto the photosensitive material, and so-called direct exposure. Exposure is mainstream.
On the other hand, in recent years, an image recorded on a film is photoelectrically read and the read image is converted into a digital signal, and then subjected to various image processing to obtain image data for recording. The photosensitive material is exposed with modulated recording light and output as a print, and the image data is output as an image file to a recording medium such as a CD-R or HD (hard disk).
[0003]
According to such digital processing, an image photographed on a film is read and image processing is performed as digital image data, so that not only color and density can be corrected very suitably, but also gradation correction and sharpness processing ( Image processing that is basically impossible with a normal direct exposure printer, such as sharpness correction, can be performed to obtain a high-quality image.
[0004]
By the way, in the output of an image using a film as a document, a major factor causing image quality degradation is image defects (hereinafter referred to as dust) caused by foreign matters such as dust and dust attached to the film, scratches on the film formed by friction, and the like. There is a wound).
In a conventional direct exposure printer, the film is manually cleaned by an operator, and an image (film) is corrected with a color material, thereby outputting a print in which such dust scratches are corrected. On the other hand, in the digital processing that reads the film image photoelectrically and handles it as digital image data, the image data obtained by reading is analyzed by image analysis, and is detected and corrected by image processing. can do.
[0005]
By the way, in a digital printer, as a method for detecting dust scratches on such a film, conventionally, an invisible image such as infrared light (IR light) is read and the invisible image is analyzed. The method of doing is known (refer patent document 1, patent document 2, etc.).
In other words, IR light is not absorbed by the image (pigment) photographed on the film, but is shielded / scattered by dust scratches, so that dust scratches are detected from changes in signal intensity when the film is read with IR light. be able to.
[0006]
However, while the dust scratch detection method using IR light has very good detection accuracy, it is necessary to provide an image reading function using IR light to the image reading device (scanner), which increases the cost of the device. Also, the reading speed is reduced.
In black and white films, some dyes and reduced silver that form an image may have an IR light absorbing component. Therefore, dust scratch detection by IR light cannot be used for such black and white films.
[0007]
Under such circumstances, there is a demand for a method for detecting dust scratches by a simple method using R (red), G (green) and B (blue) image data (RGB components). Various methods have been proposed.
As an example of a method for detecting dust flaws in an image by image analysis using the RGB components, a method of detecting a line segment that satisfies a condition by edge detection and detecting the flaw by using the line segment (patent) Document 3), or a method for detecting dust scratch candidate areas from the size of the difference between the smoothed image and the original image, and detecting dust scratches from the characteristics of the candidate areas such as changes in shape and color (see Patent Document 4) It has been known.
[0008]
However, there are many cases in which an image has a large number of regions having a profile similar to a dust scar, such as a patterned clothes.
For this reason, dust scratch detection by image analysis using these RGB components has a problem that many areas are erroneously determined as dust scratches in the image even though they are not originally dust scratches. If image correction (dust flaw correction) is performed based on dust flaw detection by image analysis, the image quality of the reproduced image may be deteriorated.
[0009]
[Patent Document 1]
Japanese Patent Application Laid-Open No. 06-028468
[Patent Document 2]
Japanese Patent Laid-Open No. 11-075039
[Patent Document 3]
JP 2000-074846 A
[Patent Document 4]
JP 2000-092319 A
[0010]
[Problems to be solved by the invention]
An object of the present invention is to solve the above-mentioned problems of the prior art, and an image obtained by photoelectrically reading an image taken on a film or the like, or a subject is photographed photoelectrically by a digital still camera. Photographing foreign objects such as dust and dust attached to the film, scratches attached to the film, foreign matter such as dust and dirt attached to an image sensor such as a CCD, or a subject from the image, or a subject. An object of the present invention is to provide an image defect detection method capable of detecting an image defect caused by a foreign substance such as dust or dust attached to an image sensor such as a CCD or the like with high accuracy without erroneous detection.
[0011]
[Means for Solving the Problems]
In order to achieve the above object, the present invention provides an image data obtained by photoelectric reading of a manuscript or photoelectric photographing of a subject. The image defect candidate of the image reproduced by the image data caused by the foreign matter attached to the reading system or the foreign matter attached to the photographing system of the subject is detected, and the density is determined according to the density at which the image defect candidate exists. An image defect detection method characterized in that image defect candidates existing at locations exceeding a predetermined state are removed and the rest are used as image defects of an image reproduced by the image data. Is to provide.
[0012]
In such an image defect detection method of the present invention, as the removal of the image defect candidate existing in a place where the density exceeds a predetermined state, the image defect candidate is enlarged and then subjected to a reduction process. It is preferable to perform at least one of the removal of image defect candidates that have caused a predetermined variation in the image of the image and the removal of image defect candidates having a predetermined number of image defect candidates within a certain range including itself. Further, it is preferable to create a luminance component image from the image data, and to detect the image defect candidate using the luminance component image, and when the image data is image data of a positive image, It is preferable to perform an image inversion process before the detection of the image defect candidate, and further, the size of the image defect in the processed image is determined before the detection of the image defect candidate. So that a constant range is preferably performed scaling processing of the image at a preset magnification in accordance with the image size.
Further, when the image defect detection method of the present invention is carried out, the display of the detection result of the dust scar on the operation screen for dust scratch detection or further correction is displayed larger than the detection result of the dust scratch. Is preferred.
[0013]
DETAILED DESCRIPTION OF THE INVENTION
Hereinafter, an image defect detection method of the present invention will be described in detail based on a preferred embodiment shown in the accompanying drawings.
[0014]
FIG. 1 shows a block diagram of an embodiment of a digital printing system for implementing the image defect detection method of the present invention.
A digital printing system (hereinafter referred to as a printing system) 10 shown in FIG. 1 photoelectrically reads an image taken on a film F, or a subject is a digital still camera (DSC, hereinafter referred to as a digital camera). 15 is used to output the image data (image file) of the image to be reproduced on the print or the recording medium such as a CD-R which has been photoelectrically photographed and reproduced on the recording medium. (Image reading device) 12, image processing device 14, and printer 22. Note that the image processing device 14 is provided with a connection port 16 for connecting the digital camera 15 in order to read image data from the memory of the digital camera 15, or image data of an image taken by the digital camera 15 is recorded. A drive 17 for reading image data from the recorded recording medium is connected. In addition, a display 18 and an operation system 20 (a keyboard 20a and a mouse 20b) are connected to the image processing apparatus 14.
[0015]
The scanner 12 is a device that photoelectrically reads an image photographed on the film F. As shown in the conceptual diagram of FIG. 2, the scanner 12 has a light source 24, a driver 26, a diffusion box 28, a carrier 30, an imaging lens unit 32, and a reading unit. 34, an amplifier 36, and an A / D (analog / digital) converter 38.
[0016]
In the scanner 12 of the illustrated example, the light source 24 uses an LED (Light Emitting Diode), and R (red) light, G (green) light, and B (blue) for reading an image photographed on the film F. It consists of LEDs that emit light. The light source 24 is driven by a driver 26, and R, G, and B lights are sequentially emitted.
The light emitted from the light source 24 enters the diffusion box 28. The diffusion box 28 makes light incident on the film F uniform in the film surface direction.
[0017]
The carrier 30 intermittently conveys the film F and sequentially conveys / holds each image (each frame) photographed on the film F to a predetermined reading position. A plurality of types of carriers 30 corresponding to the film size and the like are prepared, and are configured to be detachable from the main body of the scanner 12.
In the illustrated example, the carrier 30 includes a pair of transport rollers 40 a and 40 b that transport the film F in the longitudinal direction, and a mask 42 that regulates the reading area of each frame at a predetermined reading position. Have
[0018]
The imaging lens unit 32 focuses the projected light of the film F on a predetermined position of the reading unit 34.
The reading unit 34 uses an area CCD sensor as an image sensor for photoelectrically reading the image carried on the film F and photoelectrically reads the film F (its projection light) and is regulated by the mask 42 of the carrier 30. The entire surface of one frame is read (image reading by planar light irradiation).
[0019]
In such a scanner 12, when reading the film F, the film F is conveyed by the carrier 30, and the frame to be read is conveyed to the reading position.
Next, under the action of the driver 26, for example, the R LED of the light source 24 is driven to emit R light. The R light is made uniform in the surface direction of the film F by the diffusion box 28, then enters the reading position, enters and passes through the film F (the reading position frame), and the image photographed on this frame Becomes the projection light carrying This projection light is imaged at a predetermined position (light receiving surface of the area CCD sensor) of the reading unit 34 by the imaging lens unit 32, and the R image of this frame is photoelectrically read.
[0020]
Next, the G and B LEDs of the light source 24 are sequentially emitted to read the G image and the B image in the same manner, and the reading of this frame is completed. Note that the size of the image read by the scanner 12 (number of read pixels) differs depending on the size of the print (number of output pixels) instructed to be output.
When the reading of one frame is completed, the carrier 30 conveys the film F and conveys the next frame to be read to the reading position.
[0021]
An output signal from the reading unit 34 is amplified by an amplifier 36, converted into a digital image signal by an A / D converter 38, and output to the image processing unit 14 (data correction unit 44).
[0022]
In the printing system 10, a fine scan for reading an image at a high resolution for outputting a print or the like for one frame, a fine scan reading condition, and an image processing in the image processing device 14 (image processing unit 50) are usually performed. In order to determine the conditions, image scanning is performed twice, that is, pre-scanning, which is image scanning at a low resolution, performed prior to fine scanning.
In this case, normally, the output signals of the pre-scan and fine scan are basically the same data except that the resolution and the output level are different.
[0023]
As described above, the digital image signal output from the scanner 12 is output to the image processing device 14. FIG. 3 shows a conceptual diagram of the image processing apparatus 14.
As illustrated in FIG. 3, the image processing apparatus 14 includes a data correction unit 44, a Log converter 46, a frame memory 48 (hereinafter referred to as FM 48), an image processing unit 50, and a data conversion unit 52. Further, as described above, the connection port 16 of the DSC 15 is provided in the image processing apparatus 14, and the drive 17, the display 18, and the operation system 20 are connected thereto. Note that the image processing apparatus 14 may have a processing path for branching downstream from the Log converter 46, processing prescan data, and displaying a simulation image for verification.
The image processing apparatus 14 may be configured by combining, for example, a computer having a CPU (Central Processing Unit) and a memory, software corresponding to the computer, or further hardware.
[0024]
The data correction unit 44 is a part that applies predetermined corrections such as DC offset correction, dark correction, and shading correction to the R, G, and B output data output from the scanner 12.
The Log converter 46 performs Log conversion on the output data processed by the data correction unit 44 by using, for example, an LUT (Look Up Table) or the like to obtain digital image (density) data (image information).
The R, G, and B image data converted by the Log converter 46 are stored in the corresponding FM 48, respectively. Further, in the illustrated example, when processing each image data of the supplied image (image file) from the digital camera 15 via the connection port 16 or via the drive 17 from a recording medium or the like, an example is given. Are supplied from the corresponding I / F (interface) 54 and 55 to the FM 48 of each color and stored therein.
[0025]
The image processing unit 50 performs enlargement / reduction processing (electronic scaling processing), color / density correction, gradation conversion, sharpness processing (sharpness enhancement), image density dynamic range compression (image data) on the image data stored in the FM 48. This is a part that performs various image processing such as image correction processing such as dodging effect by processing) and special processing such as soft focus processing and cross filter processing.
These image processes may be performed by a known method using a look-up table (LUT), matrix calculation, filter processing, and the like. The image processing conditions are basically images using pre-scan image data. Set by analysis.
[0026]
In addition, when the image processing unit 50 is instructed to perform image defect correction (dust scratch correction), the image processing unit 50 uses the image defect detection method of the present invention to detect scratches on the film F and dust attached to the film F. An imaging system such as a digital camera 15 for capturing a foreign object such as dust or dust, a foreign matter such as dust or dust adhering to the light receiving surface of the area CCD sensor of the scanner 34, particularly a reading unit 34, or a subject such as a CCD. Detects image defects (hereinafter referred to as dust scratches) caused by foreign matters such as dust and dust attached to the light receiving surface of the image sensor, and further detects dust scratches detected by this dust scratch detection and inputs by the operator. Correct the specified dust flaw. In the following description, scratches on the film F and foreign matters such as dust and dirt adhering to the film F are used as representative examples of dust scratches, but it goes without saying that the present invention is not limited to this.
Note that the timing of dust scratch detection and correction in the image processing unit 50 is not particularly limited, and may be performed prior to other image processing or after all other image processing is completed. It may be incorporated in the middle of image processing (for example, after performing color density-related image processing and before image structure-related processing). Further, if the image structure characteristics match, processing may be performed using images in different states for detection and correction.
[0027]
Hereinafter, an example of the dust scratch detection method of the present invention will be described with reference to the flowcharts of FIGS. 4 and 5.
In this example, as a preferred mode, first, electronic scaling processing of image data is performed to normalize the image size. This standardization of the image size is a process for measuring a reduction in the amount of image data to be processed and making it possible to use uniform dust flaw detection parameters for images of all sizes.
[0028]
For example, if the output image is 2L size, the read image size is about 2100 pixels × 1500 pixels, and if the output image is wide four size, the read image size is about 3000 pixels × 4500 pixels, as described above. The image reading size (the number of read pixels) in FIG. 12 differs depending on the output image size (the number of output pixels) such as the print size.
On the other hand, the size (thickness) of foreign matters such as dust adhering to the film F and scratches entering the film F is relatively uniform regardless of the film size and film type.
[0029]
Therefore, even in the same frame, if the size of the read image is different, the size of the dust scratch (the number of pixels of the dust scratch) is different on the read image. Therefore, in order to detect dust flaws with high accuracy, it is necessary to set dust flaw detection parameters according to the image size.
In addition, when the image size is large, there is a problem that processing takes time.
[0030]
On the other hand, in the dust scratch detection of the illustrated example, as a preferred mode, the size of the image to be processed is reduced, and the dust scratch on the processed image is set in a predetermined range according to the read image size. The image is electronically scaled to normalize the image size.
Accordingly, dust scratch detection can be performed with a common parameter corresponding to various image sizes, and dust scratch detection can be quickly performed even with a large size image.
[0031]
For the electronic scaling factor (enlargement / reduction ratio) for standardization, for example, a standard image size (reference size) in dust scratch detection is determined, and the size of dust scratches on the processed image (image on which dust scratch detection is performed). However, an electronic scaling factor that is the same as the reference size may be determined as appropriate.
Here, the electronic scaling process for standardization may be performed for all cases except the reference size. However, if the image size is smaller than the reference size, enlargement leads to an increase in processing time, which is disadvantageous in terms of productivity. Therefore, it is also preferable not to perform such standardization of the image size to a size equal to or smaller than the reference size. Whether or not to standardize the image size for a size equal to or smaller than the reference size may be determined as appropriate in consideration of the required processing speed and detection performance. In addition, for the size equal to or smaller than the reference size, the presence or absence of normalization may be selected.
[0032]
Next, a Y component image (luminance component image) is created from the R, G, and B images. There are no particular limitations on the method of creating the Y component image, and various known methods can be used. For example, the following method may be used.
Y = 0.3R + 0.59G + 0.11B
[0033]
Further, when the image that is the original is a positive image (film F is a reversal film), gradation inversion is performed to obtain a Y component image similar to the negative image. The gradation inversion method is not particularly limited and may be a known method.
That is, in this example, as a preferable mode, image types such as a negative image, a positive image, a color image, and a monochrome image are obtained by normalizing the image size, extracting the Y component, and further reversing the gradation of the positive image. Regardless of the image size, regardless of the image size, it is possible to detect dust flaws with high accuracy using common processing and parameters for all types of images.
[0034]
An example of an image in this state is shown in FIG. This image is an image of a parade of illuminated floats.
Using this image as image 1, an opening process of a morphological filter is performed on image 1 to create image 2. Next, a difference image is created by subtracting image 2 from image 1.
[0035]
Next, the threshold value set according to the dust scratch detection level is compared with the average value of the difference image. If the difference between the threshold value and the average value of the difference image is equal to or less than a predetermined value, the threshold value is adjusted. .
[0036]
When foreign matter is attached to the film F, the light in that region is shielded, so that the image density becomes high. If the film is scratched, the light in that area is diffused, and the image density is similarly increased. Therefore, in the above-described difference image, the image data other than the predetermined range of image data (below the predetermined luminance) cannot be considered as dust scratches in terms of image density. In the dust scratch detection in the illustrated example, first, an area considered not to be a dust scratch is deleted from such a difference image using a threshold value.
This threshold value is determined in advance according to the “detection level” on the operation screen shown in FIG. Further, the “detection level” can be adjusted by an operator who operates the printing system 10. Note that “detection level” is high means that dust scratches are strongly detected (even if the possibility is lower, it is determined that dust scratches are present), that is, the threshold value is low.
[0037]
Here, in the difference image obtained by the “image 1-image 2” described above, if the document is an image such as a dark positive image, a very large number of high-luminance regions are generated in the difference image. However, it is natural that all of the high luminance areas are not dust scratches.
That is, in such an image, the threshold according to the detection level is not necessarily appropriate, and as a result, the area detected as a dust scratch candidate increases, and the subsequent processing time becomes very long. There is a problem that processing efficiency decreases.
[0038]
Here, since a difference image such as a dark positive image has a very high average value, it is determined whether or not the threshold value is appropriate for the image by comparing the difference between the average value and the threshold value. If the threshold is too low compared to the average value, the threshold can be adjusted to shorten the time and perform efficient processing.
[0039]
The difference between the threshold for adjusting the threshold and the average value of the difference images is not particularly limited, and may be set as appropriate according to the required processing time. According to the study by the present inventors, it is preferable to adjust the threshold when the threshold is not more than three times the average value of the difference images in terms of processing time and detection accuracy. That is, when [threshold] / [average of difference images] ≧ 3 is not satisfied, it is preferable to adjust the threshold.
Further, there is no particular limitation on the adjustment of the threshold, and a threshold that satisfies the condition may be set according to the difference between the threshold for adjusting the threshold and the difference image. For example, in the above-described example, a method of [adjusted threshold] = [average difference image × 3] + [threshold according to the detection level] is exemplified.
[0040]
In the present invention, when the dust scratch detection level is higher than a predetermined level, that is, when the threshold value is lower than the predetermined value, in order to save processing time, “image 1-image 2”. After the difference image is created by “”, another process may be entered after that (that is, after branching from the comparison judgment between the threshold value and the difference image) (arrow b in FIGS. 4 and 8). reference).
This will be described in detail later.
[0041]
Next, the difference image is binarized with a threshold value corresponding to the above-described detection level, and an image 3 in which image defect candidates are extracted based on the threshold value is created. Further, the isolated point is deleted from the image 3.
The isolated point is a very small area although it remains as a dust flaw candidate in the image 3. As described above, the size of the dust scratches on the film F is known to some extent. Therefore, in the image 3 from which the dust scratch candidates are extracted, a region (isolated point) that is smaller than the size that can be considered as dust scratches is deleted, and the subsequent processing is performed, thereby quickly detecting dust scratches. be able to.
The size of the isolated point may be determined as appropriate in accordance with the above-mentioned standardization of the image. However, in terms of accuracy and efficiency, an area of 8 pixels or less, particularly a single pixel existing in the image 3 is selected. It is preferable to delete as an isolated point.
[0042]
By deleting the isolated points, dust flaw candidates (image defect candidates) are detected (see FIG. 6B), and dust flaws corresponding to the existence density, which is a characteristic point of the present invention, are obtained in subsequent processing. Candidate removal is performed and dust flaws in the image are detected.
[0043]
According to the study of the present inventor, in the detection of dust scratches using R, G, and B image data, it is easy to mistakenly detect dust scratches in places with patterns such as clothes and the density of human hair. , Windows of buildings, electrical illumination as shown in FIG. 6, etc., and in such places, many dust scratch candidates are densely detected. In other words, a place where dust scratch candidates are detected at a high density is highly likely to be a false detection.
Also, a noticeable problem in image quality when reproduced as a visible image is dust scratches in areas with little fluctuation, such as a uniform density area (solid area), and the image fluctuates on the contrary. In many cases, there are dust scratches in a place (place where the image is busy), and even if this is reproduced as a visible image, it is not noticeable and there is no problem in image quality.
[0044]
For this reason, after detection of dust scratch candidates, depending on the density at which dust scratch candidates exist, high density portions are removed as not being dust scratches, and only dust scratch candidates present at low density locations are removed. By determining it as a flaw, it is possible to reliably detect a dust flaw that leads to a large deterioration in image quality, and to greatly reduce false detection. Thus, for example, when dust damage correction is performed with the print system 10 as shown in the drawing, the processing speed and processing efficiency can be improved, that is, a system with high productivity can be realized.
[0045]
In the present invention, there are no particular limitations on the method for detecting dust flaw candidates (image defect candidates), and various methods for detecting dust flaws using R, G, and B images can be used. For example, dust scratch candidates may be extracted by a method using edge detection as described in Patent Document 3 and a method using a difference between a smoothed image and an original image as described in Patent Document 4. In addition to this, it is also possible to use a method for detecting dust scratches by performing image analysis and utilizing pixel continuity, color balance with surrounding pixels, and the like.
[0046]
In the dust scratch detection shown in the flowcharts of FIGS. 4 to 5, when isolated points are deleted from the image 3 as described above, first, labeling is performed on each dust scratch candidate existing in the image 3 from which the isolated points have been deleted. Processing (numbering) is performed, and this is stored as image A.
Further, an image 4 is generated by performing a closing process of the morphological logic filter on the image 3, and a dust scratch candidate existing in the image 4 is labeled and stored as an image B.
[0047]
Next, the images A and B are analyzed (FIG. 5), and dust scratch candidates having different areas of the image A and the image B by a predetermined value or more and a predetermined size or more (n pixels or more) in the image A are detected. If a predetermined number (m) or more are combined in B, the corresponding dust scratch candidate is deleted. That is, the dust flaw candidate that satisfies the above conditions is a place where the density of dust flaw candidates is high and is determined not to be a dust flaw, and is deleted from the image 3.
The deleted image is saved as image 5.
[0048]
The closing process of the morphologic filter is, so to speak, a process of enlarging an image and then reducing the enlarged image at a magnification to be restored.
Accordingly, in a place where dust scratch candidates are low density, the dust scratch candidates do not change at all before and after the closing process, or the change amount is small even if the area or the like changes. On the other hand, in places where dust scratch candidates are dense, neighboring objects are joined together by enlargement and are reduced from this state, so that a large change occurs in the dust scratch candidates due to the closing process.
[0049]
Therefore, it is determined whether or not the dust flaw candidate exists in a place where the dust flaw candidate exists at a high density according to the combined state of the area variation in the images A and B and the neighboring dust flaw candidate in the image B. be able to.
Further, as described above, since the size of dust scratches is known to some extent, the processing can be quickly performed by excluding dust scratch candidates whose size in image A is equal to or smaller than a predetermined value. . In the illustrated example, all the dust scratch candidates of n pixels or less are deleted as noise in the process described later.
[0050]
Here, the shape of the dust scratches is not limited to a straight line, and there are many, for example, a shape having a bent shape, a bent shape, a round shape, or the like. When closing the dust scratches having such a shape, for example, a rounded portion is crushed during enlargement, and as a result, even if there are no dust scratch candidates in the vicinity, a large change occurs in the area and shape before and after the processing. . Therefore, if only the area variation due to the closing process is determined, such a dust scratch candidate may be determined as not being a dust scratch and deleted.
On the other hand, in addition to the area change due to the closing process, by adding a determination that “m or more pieces are combined in the closing image (B image)”, not only the deletion of dust scratch candidates existing in a high-density place, It is possible to prevent such dust flaw candidates from being erroneously deleted, and to detect dust flaws with higher accuracy.
[0051]
There is no particular limitation on the area variation of the image A and the image B to be excluded from the dust scratch candidates, and it may be determined as appropriate according to the standard size for normalization of the image size. Also, there are no particular limitations on the n pixels that are excluded from dust scratch candidates as noise, and the number m that is the number of connections determined to be high density, and may be appropriately determined according to the reference size and the like in the same manner. .
If the standard size for standardization is 2L (about 2100 pixels × 1500 pixels), for example, the image A and the image B have an area that is twice or more different from each other, and there are dust scratch candidates of 8 pixels or more. It is preferable to delete four or more dust flaw candidates that are combined.
[0052]
Next, a labeling process is performed on the image 5 to delete dust scratch candidates according to the number (presence density) of dust scratch candidates per predetermined area, and to delete noise (noise). This process will be described with reference to the flowchart of FIG.
[0053]
As shown in FIG. 7, first, the labeled image 5 is scanned (scanned) from the pixel [0, 0] to detect a pixel in the dust flaw candidate of the image 5 (of the labeled pixel). detection). This scanning is performed to the end of the image.
When a pixel in the dust scratch candidate is detected, it is first checked whether or not the number of pixels of the dust scratch candidate is n pixels or less. If the pixel is equal to or less than n pixels, the dust scratch candidate is deleted (this label Deleted). Note that the n pixels are set according to the standardized reference size and the like as described above, and as an example, the n pixels are 8 pixels. Further, if necessary, the number of pixels determined by the closing process may be different from the number of pixels.
[0054]
On the other hand, if the number of dust scratch candidates exceeds n pixels, the detected pixel is used as a target pixel, and a predetermined area centering on the target pixel is searched to detect a dust scratch candidate existing therein. . The predetermined area in this case is not particularly limited, and may be appropriately determined according to the standard size for standardization. However, as long as the size is 2L, as an example, 100 pixels × 100 pixels Is exemplified.
From this search result, if the label is different from the dust flaw candidate in which the target pixel exists in the area, and there are more than a predetermined number of dust flaw candidates of a predetermined size, they exist in a high-density place. It is determined that it is a dust flaw candidate to be deleted, and a dust flaw candidate in which the pixel of interest exists is deleted (this label is deleted). The deletion criterion is not particularly limited, and may be appropriately determined according to the standard size for standardization. However, as an example of the 2L size, 9 to 100 pixels can be used. When two or more things exist, deleting is exemplified.
[0055]
When the determination is made, scanning is performed to go to a pixel in the next dust scratch candidate, and it is determined whether or not the label of the dust scratch candidate is checked (determined above). In Y), going to the next dust wound candidate is repeated.
If this dust scratch candidate label is a label for a dust scratch candidate that has not been checked (N), the process returns to the top to detect the number of dust scratch candidate pixels in which this pixel exists, and within a predetermined area. Search for.
[0056]
The above routine is repeated, and when the end of the image is reached, the process is terminated.
Thereby, only dust scratch candidates in places where the number of pixels is n or more and the density of existence is low can be detected as dust scratches (see FIG. 6C). As described above, this extraction result is highly accurate with few false detections.
[0057]
Basically, the dust scratch detection of the present invention is completed by the above operation.
Here, since the dust scratch usually has a light and shade, the dust scratch area detected by the above dust scratch detection may not indicate the entire dust scratch. For this reason, if only the detected area is corrected as dust scratches, correction may actually be performed using image data of the area that is dust scratches. In this case, correct correction should be performed. I can't.
Therefore, in the illustrated image processing unit 50, in order to enable highly accurate dust scratch correction, the dust scratch detected by the above dust scratch detection is expanded, and this is used as a final dust scratch detection result. And
[0058]
There is no particular limitation on the expansion of dust scratches at this time, and an example is expansion that expands the pixels to 8 pixels surrounding the periphery, or expansion that further expands the surrounding pixels to 16 pixels.
In addition, for example, the degree of expansion of dust scratches at this time may be changed according to the dust scratch correction level described later, such as the greater the dust scratch correction strength, the greater the expansion.
[0059]
As described above, in dust scratch detection according to the present invention, if the dust scratch detection level is higher than a predetermined value, that is, if the threshold is lower than a predetermined value, the processing time is saved. For this reason, once the difference image is created, another process may be performed thereafter.
[0060]
Hereinafter, this process will be described with reference to the flowchart of FIG.
When the difference image is created by performing the process of “image 1-image 2” as shown in FIG. 4 (see arrow b in FIG. 4 and FIG. 8), in this process, as shown in FIG. The difference image is binarized with a threshold value corresponding to the detection level, and an image 3 in which dust wound candidates are extracted based on the threshold value is created. Next, the isolated point is deleted from the image 3.
[0061]
Next, the image 4 from which the isolated points have been deleted is subjected to a morphological logic closing process to create an image 4. Further, a labeling process is performed on the image 4.
A dust scratch candidate having a predetermined number of pixels is extracted from the image 4 that has been subjected to the labeling process, and this is used as an image 5. Note that the number of dust scratch candidate pixels extracted from the image 5 is not particularly limited, and may be determined as appropriate according to the standard size in the standardization. If the reference size is the 2L size, 8 to 1600 pixels are exemplified as an example.
[0062]
When the extraction is completed, the AND of the image 5 and the image 3 is taken. That is, from the image 3, only dust scratch candidates existing in both the image 5 and the image 3 are extracted. Next, the extracted dust flaw candidate is labeled.
Thereafter, the labeling-processed image is subjected to processing of “removal according to density and noise removal” in the flowcharts of FIGS. 4 to 5 (see arrows c in FIGS. 5 and 8). This process is performed to end the detection of dust scratches.
[0063]
A high detection level = low threshold means that a large number of dust scratch candidates are detected, and the above-described method takes time. On the other hand, according to the processing shown in FIG. 8, even when such a threshold is low, the processing time can be shortened and dust scratches can be detected with good detection accuracy. The dust scratch detection level (threshold value) entering this process may be appropriately set according to the target processing time, dust scratch detection accuracy, and the like.
[0064]
In the dust scratch detection shown in FIGS. 4, 5 and 8, the closing process of the morphologic filter is preferable as a preferred mode that can improve the detection accuracy in order to remove dust scratch candidates in a high-density place. Although both deletion of candidates by (enlargement and reduction processing) and deletion of candidates using the number of dust flaw candidates within a predetermined range are performed, the present invention is not limited to this, and either one is removed You may only do.
[0065]
As described above, when the dust scratch correction is instructed, the image processing unit 50 corrects the dust scratch detected by such dust scratch detection or the dust scratch input by the operator.
There are no particular limitations on the dust scratch correction method, and various known methods such as interpolation using the peripheral pixel information of the dust scratch area can be used. In the illustrated example, interpolation using this peripheral pixel information is possible. To correct dust scratches.
Further, as will be described later, in the illustrated example, the level of dust flaw correction (correction strength) can be changed. As an example, the dust scratch correction level may be adjusted according to the size of the area used for interpolation. Furthermore, the dust scratch correction level is corrected according to the amount of expansion during dust dust detection for dust scratch detection. The intensity may be changed. In both cases, the wider the width, the higher the correction level.
[0066]
Hereinafter, an example of such dust flaw correction operation will be described. This operation may be performed by a known method using a GUI (Graphical User Interface).
As described above, when the image data is stored in the FM 48 and the image processing unit 50 reads out the image data, if the dust scratch correction is instructed at a predetermined timing, the print system 10 enters the dust scratch correction mode. Then, a dust flaw correction screen as shown in FIG.
[0067]
In FIG. 9, the image displayed on the left (hereinafter referred to as an original image) is subjected to dust flaw correction. On the other hand, a similar original image is also displayed on the right side. This original image is displayed with a dust scratch detection result, a pointer for performing various operations, a range designation frame, and the like superimposed. Note that in the initial stage, the detection result of dust and the like is not displayed, and only the original image is displayed.
In addition, the up and down and left and right scroll bars for changing the display area are displayed only on the left image, but both images move in conjunction with the movement of the scroll bar.
[0068]
In the illustrated example, the image is rotated by 90 ° by pressing the “rotate 90 degrees” button. The right button rotates right and the left button rotates left.
Further, the image is enlarged stepwise at a predetermined magnification by pressing the “zoom-in” button, and the image is enlarged and displayed by pressing the “region zoom” button and clipping with the mouse 20b. Further, in a state where the image is enlarged, the image is reduced stepwise by a predetermined magnification by pressing the “zoom out” button, and the entire image is displayed by pressing the “entire display” button.
Further, normally, on this operation screen, an image for display display is displayed in order to ensure operability, but an image close to an image reproduced for printing is displayed by pressing the “faithful reproduction” button.
[0069]
When the “automatic detection” button is pressed by the operator, the image processing unit 50 performs dust scratch detection shown in FIGS. 4 to 5 (or FIG. 8), and the original image is displayed as shown in FIG. The detection result of dust scratches is displayed over the original image on the right side. In the illustrated example, it is assumed that seven dust scratches are detected.
Here, the dust scratches on the operation screen are preferably displayed slightly larger than the final detection result shown in FIG. 5 so that fine dust scratches can be identified well or sufficiently.
[0070]
As described above, the dust scratch detection level (threshold) can be adjusted by adjusting the “detection level” slider bar prior to the automatic detection instruction. Value (10 in this example). When the “level save” button is pressed, the set detection level is saved, and the next dust scratch detection starts at this detection level.
[0071]
For example, the dust scratch detection result is initially highlighted by being colored M (magenta), but when the “change detection color” button is pressed, the dust scratch detection result is displayed in a different color. For example, C (cyan) → Y (yellow) → M is sequentially changed.
Thereby, it is possible to reliably identify dust scratches regardless of the background or the like.
[0072]
In the illustrated example, it is possible to perform various operations by specifying a range for the result of such dust scratch detection.
[0073]
In the illustrated example, a rectangular range specification using the mouse 20b and a manual range specification are set, and the rectangular range specification using the mouse 20b includes the following two types of range specification. Is set.
By pressing the upper button in the “range specification” button on the GUI of the display screen, as shown in FIG. 10A, the mouse 20b is used to select from any position designated (clicked) on the screen. By dragging in an arbitrary direction, it is possible to specify an arbitrary rectangular range whose diagonal line is the dragged line.
On the other hand, when the button below the GUI on the display screen is pressed, dragging in an arbitrary direction from an arbitrary position designated on the screen in the same manner as shown in FIG. Draw a rectangle of predetermined width with sides extending in the direction. The drawing of the rectangle ends when the drag ends, and then the range is specified with an arbitrary rectangle in the oblique direction by adjusting the width (direction orthogonal to the drag direction) by clicking the right button of the mouse or the like. By having this range designating method, it is possible to suitably designate the dust flaws entering obliquely without including unnecessary areas.
[0074]
On the other hand, manual designation using a pen tool, an eraser tool, a line segment tool, or the like can be performed by pressing a “manual designation” button. The operation method using these tools may be a known method. In the case of manual designation, the displayed dust scratch itself may be designated instead of the range designation.
Here, when the “manual designation” button is pressed, the image display on the operation screen is an enlarged display of only the image on which dust flaws are displayed (the right side of the two-image display).
[0075]
Of course, such range designation can be performed at a plurality of locations for one process, and different types of range designation can be performed for one process. is there.
[0076]
As an operation for the automatic detection result, for example, when detecting dust scratches that could not be detected correctly (detecting dust scratches that were considered to be dust scratches from the image), such range specification, After the detection level is adjusted, the “auto detection” button is pressed again. In response to this, the image processing unit 50 performs dust scratch detection at the readjusted detection level only for that region.
[0077]
In addition, when deleting the erroneously detected dust scratch (the portion detected as a dust scratch although it is not a dust scratch), the “cancel” button is pressed after the same range is designated. Thereby, the image processing unit 50 deletes dust scratches detected within the designated range. Note that erasure of erroneously detected dust scratches may be performed by performing automatic detection with the detection level set to “0” after specifying the range of the erroneous detection area.
Alternatively, by designating only an area where detection was correct and issuing an instruction, dust damage in the area not specified may be deleted as a false detection.
[0078]
After the dust scratch detection as described above is performed, when the “correction” button is pressed by the operator, the image processing unit 50 performs the dust removal by the interpolation using the image data of the peripheral region of the dust scratch as described above. Perform scratch correction. In addition, an image subjected to dust flaw correction (that is, a dust flaw correction result) is displayed on the right side of the screen.
In response to pressing of the “Correction” button, the “Correction” button transitions to the “Confirm” button, and the “Detection level” slider bar transitions to the “Correction level” slider bar.
[0079]
After the correction, when the correction level is adjusted by the slider bar, the “confirm” button is changed to the “correction” button again. When the “Correction” button is pressed in this state, the image processing unit 50 performs correction again with the adjusted correction level. Note that the re-correction is basically performed on the original image.
In addition, by pressing the “correction initialization” button, the dust-scratch-corrected image is returned to immediately before entering the dust-scratch correction.
Further, such re-correction and correction initialization may be performed only for the specified range in accordance with the above-described range specification.
[0080]
In the print system 10 shown in the drawing, when the “trial print” button on the operation screen is pressed during the dust / scratch correction operation, a print of the dust / scratch correction image at that time is performed accordingly (output of the test print). . Thereby, it becomes possible to confirm the correction state of dust scratches by actual printing.
The print output will be described in detail later.
[0081]
As described above, dust flaw correction is performed, and when it is determined that dust flaw correction is appropriate, the operator presses a “confirm” button. In response to this, the image correction unit 50 determines dust flaw correction for this image. Further, when the “output-end” button is pressed, the next processing is started for this image, and the dust flaw detection mode is ended.
Here, in the present invention, it is preferable that the image that has been “determined” after the dust flaw correction is completed is stored without being erased. As a result, for example, after a (finished) print is output, even if a dust flaw is left uncorrected or a correction error is found, additional correction can be performed from the stored image, thereby improving processing efficiency. Processing time can be shortened.
[0082]
If the “initialize all” button is pressed during the above process, all the processes for this image are canceled and the process returns to the initial state before the dust scratch detection.
Similarly, when the “cancel” button is pressed, all the operations are canceled, and the printing system 10 returns to a mode in which dust flaw correction is not performed.
Further, in the print system 10 shown in the drawing, the number of prints of the image subjected to dust flaw correction can be set on this operation screen by using the “number of sheets” up / down buttons.
[0083]
The image data processed by the image processing unit 50 is processed by the data conversion unit 52. The data conversion unit 52 is a part that converts image data into image data corresponding to image output by the printer 22 using, for example, a three-dimensional LUT (3D-LUT).
[0084]
The printer 22 is a known color printer. For example, a photosensitive material such as photographic paper is two-dimensionally scanned and exposed by a light (laser) beam modulated according to supplied R, G, and B image data, and a latent image is recorded. An example is a printer in which a material is subjected to a development / fixing / water-washing wet development process to visualize a latent image, and then dried and output as a print.
In the print system 10 of the present invention, in addition to print output, the image data processed by the correction unit 54 is converted into, for example, a JPEG image file and output to a storage medium such as a CD-R. Also good.
[0085]
The image defect detection method of the present invention has been described in detail above. However, the present invention is not limited to the above-described embodiments, and various improvements and modifications may be made without departing from the gist of the present invention. Of course.
[0086]
For example, in the above example, the image defect is detected over the entire input image, but the image defect may be detected only in the area corresponding to the output image, or the designated area. Only image defect detection may be performed, and any one of image defect detection in the entire area / output image area / designated area may be selected.
In the illustrated example, the image defect is detected using the fine scan data by the scanner 12, but the present invention is not limited to this, and if possible in terms of resolution or the like, focus on productivity. Then, the image defect may be detected using the prescan data.
Further, in the illustrated example, image defects due to dust scratches on the film F (transparent document) are detected. However, the present invention detects image defects from image data obtained by photoelectrically reading a reflective document. May be. As described above, a reading system that photoelectrically reads a film F (transmission original) or a reflection original, particularly an image defect such as dust scratches attached to a light receiving surface of an image sensor such as a CCD sensor, or a digital camera that photographs a subject ( Needless to say, image defects such as dust scratches adhering to the light receiving surface of an image sensor such as a DSC), particularly an image sensor such as a CCD sensor, may be detected.
[0087]
【The invention's effect】
As described above in detail, according to the present invention, the final image defect is determined by deleting the erroneously detected image defect candidate according to the density of the detected image defect candidates. In addition, it is possible to reliably detect image defects existing in a solid area or the like, which is a big problem when an image is reproduced, and to efficiently perform highly accurate image defect detection with few false detections.
[Brief description of the drawings]
FIG. 1 is a block diagram of an embodiment of a printing system that implements an image defect detection method of the present invention.
FIG. 2 is a conceptual diagram of an embodiment of a scanner of the printing system shown in FIG.
FIG. 3 is a block diagram of an embodiment of the image processing apparatus of the print system shown in FIG.
FIG. 4 is a flowchart of an example of an image defect detection method of the present invention.
FIG. 5 is a flowchart of an example of the image defect detection method of the present invention, which is a continuation of FIG.
FIGS. 6A, 6B, and 6C are reference diagrams for explaining the image defect detection method of the present invention. FIGS.
FIG. 7 is a flowchart for explaining the flowchart shown in FIG. 5;
FIG. 8 is a flowchart of another example of the image defect detection method of the present invention.
FIG. 9 is a conceptual diagram of an operation screen for image defect correction in the printing system shown in FIG.
FIGS. 10A and 10B are conceptual diagrams for explaining a range specifying method on the operation screen shown in FIG. 9;
[Explanation of symbols]
10 (Digital) printing system
12 Scanner
14 Image processing device
15 Digital (still) camera
16 connection port
17 drive
18 display
20 Operation system
22 Printer
24 Light source
26 Driver
28 Diffusion box
30 career
32 Imaging lens unit
34 Reading unit
36 amplifiers
38 A / D converter
40 Conveying roller pair
42 Mask
44 Data correction unit
46 Log Converter
48 FM (frame memory)
50 Image processing section
52 Data converter
54,55 interface

Claims (5)

From image data obtained by photoelectric reading of a document or photoelectric shooting of a subject, the document is damaged and the foreign matter attached to the document, or further, the foreign matter attached to the document reading system, or the subject Detecting image defect candidates in the image reproduced by the image data caused by foreign matter attached to the imaging system,
According to the density at which the image defect candidate exists, the image defect candidate existing at a location where the density exceeds a predetermined state is removed, and the rest is used as an image defect of an image reproduced by the image data. Image defect detection method. As removal of image defect candidates that exist in places where the density exceeds a predetermined state,
The image defect candidate is enlarged and then reduced, the image defect candidate having a predetermined fluctuation is removed from the processed image, and the predetermined image defect candidate within a certain range including itself is removed. The image defect detection method according to claim 1, wherein at least one of removal of image defect candidates having an existing number equal to or greater than a specified number is performed. The image defect detection method according to claim 1, wherein a luminance component image is created from the image data, and the image defect candidate is detected using the luminance component image. The image defect detection method according to claim 1, wherein when the image data is image data of a positive image, an image inversion process is performed prior to detection of the image defect candidate. The image scaling processing is performed at a magnification set in advance according to the image size so that the size of the image defect in the processed image falls within a predetermined range before the detection of the image defect candidate. 5. The image defect detection method according to any one of 4 above.

Images