WO2016098323A1 - 情報処理装置、情報処理方法、及び、記録媒体 - Google Patents
情報処理装置、情報処理方法、及び、記録媒体 Download PDFInfo
- Publication number
- WO2016098323A1 WO2016098323A1 PCT/JP2015/006170 JP2015006170W WO2016098323A1 WO 2016098323 A1 WO2016098323 A1 WO 2016098323A1 JP 2015006170 W JP2015006170 W JP 2015006170W WO 2016098323 A1 WO2016098323 A1 WO 2016098323A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- image
- attribute
- information
- pixel
- input image
- Prior art date
Links
- 230000010365 information processing Effects 0.000 title claims abstract description 110
- 238000003672 processing method Methods 0.000 title claims description 5
- 230000008859 change Effects 0.000 claims abstract description 133
- 238000004364 calculation method Methods 0.000 claims description 79
- 238000000034 method Methods 0.000 claims description 40
- 230000008569 process Effects 0.000 claims description 15
- 230000000593 degrading effect Effects 0.000 claims 1
- 238000012545 processing Methods 0.000 abstract description 47
- 239000013598 vector Substances 0.000 description 54
- 230000006870 function Effects 0.000 description 29
- 238000010586 diagram Methods 0.000 description 18
- 230000000694 effects Effects 0.000 description 13
- 239000011159 matrix material Substances 0.000 description 13
- 238000007781 pre-processing Methods 0.000 description 12
- 238000012937 correction Methods 0.000 description 7
- 230000001629 suppression Effects 0.000 description 5
- 230000006866 deterioration Effects 0.000 description 4
- 230000014509 gene expression Effects 0.000 description 4
- 238000010191 image analysis Methods 0.000 description 4
- 238000012986 modification Methods 0.000 description 4
- 230000004048 modification Effects 0.000 description 4
- 239000004973 liquid crystal related substance Substances 0.000 description 3
- 230000003287 optical effect Effects 0.000 description 3
- 230000002093 peripheral effect Effects 0.000 description 3
- XEEYBQQBJWHFJM-UHFFFAOYSA-N Iron Chemical compound [Fe] XEEYBQQBJWHFJM-UHFFFAOYSA-N 0.000 description 2
- 238000003702 image correction Methods 0.000 description 2
- 238000003384 imaging method Methods 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 238000003909 pattern recognition Methods 0.000 description 2
- 238000012706 support-vector machine Methods 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 238000002939 conjugate gradient method Methods 0.000 description 1
- 238000011840 criminal investigation Methods 0.000 description 1
- 238000006731 degradation reaction Methods 0.000 description 1
- 238000006073 displacement reaction Methods 0.000 description 1
- 229910052742 iron Inorganic materials 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 239000002023 wood Substances 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T1/00—General purpose image data processing
- G06T1/20—Processor architectures; Processor configuration, e.g. pipelining
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T1/00—General purpose image data processing
- G06T1/60—Memory management
-
- G06T5/70—
-
- G06T5/73—
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N1/00—Scanning, transmission or reproduction of documents or the like, e.g. facsimile transmission; Details thereof
- H04N1/40—Picture signal circuits
- H04N1/409—Edge or detail enhancement; Noise or error suppression
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
Definitions
- the present invention relates to image processing, and more particularly to an information processing apparatus, information processing method, and recording medium related to parameter adjustment in image processing.
- An information processing apparatus that restores an image uses, for example, the following operation in order to restore a degraded image that has deteriorated due to blurring, noise, or a reduction in resolution.
- the information processing apparatus generates a temporary restored image (candidate restored image) based on a target degraded image (hereinafter also referred to as an input image) using the default parameters.
- the information processing apparatus generates an image that simulates (applies) a deterioration process such as blur (blurring effect) to the restored image.
- the information processing apparatus corrects the restored image so that the difference between the image generated using the simulation and the target degraded image is minimized.
- a solution for example, a pixel value in an image
- the information processing apparatus cannot uniquely determine a solution. Therefore, the information processing apparatus uses, for example, a constraint (for example, regularization) on the solution in order to uniquely determine the solution. That is, the information processing apparatus uses regularization to constrain the solution and uniquely determine the solution.
- the regularization used as the constraint is, for example, a constraint that suppresses a change in pixel value between adjacent pixels in the restored image.
- the information processing apparatus uniquely determines a solution as a restored image using the regularization described above (see, for example, Patent Document 1).
- the technique described in Patent Document 1 uses regularization strength in order to achieve both a clear texture area and noise suppression in a flat area. Specifically, the technique described in Patent Document 1 is based on the direction of change between adjacent pixels and the magnitude of change in pixel value for the pixels constituting the input image, and the pixel value between adjacent pixels in the restored image. The regularization strength is determined so that the difference amount of.
- restoring an image may be referred to as reconstructing the image. Therefore, the restored image may be called a reconstructed image.
- Patent Document 1 requires parameters corresponding to the amount of change in pixel value and the regularization strength in order to determine the regularization strength. For this reason, the information processing apparatus needs to receive these values from the user prior to image processing.
- the image quality desired for an image differs depending on the object included in the image, the purpose of the image, and the user of the image. For example, how much the noise of the flat portion in the image is suppressed differs depending on the use of the image and the user of the image. Therefore, a user of an information processing apparatus that processes an image needs to set an appropriate parameter value or adjust the parameter value to an appropriate value for the image processing apparatus in accordance with the purpose of the image processing. there were.
- the technique described in Patent Document 1 has a problem that it is difficult to adjust (set) parameters in image processing. Therefore, the technique described in Patent Document 1 has a problem that an appropriate restored image cannot be provided.
- An object of the present invention is to provide an information processing apparatus, an information processing method, and a recording medium that solve the above-described problems and allow a user to adjust (specify) appropriate parameters in image processing more easily. is there.
- An information processing apparatus includes an amount-of-change calculating unit that calculates an amount of change between a value of a predetermined pixel of an input image and a value of a pixel around the pixel in an input image to be processed; Attribute reliability calculation means for calculating attribute reliability, which is the reliability of the attribute of the pixel of the input image, based on the attribute that is the property of the pixel of the specified area in the input image and the amount of change; Based on the image quality information that is the image quality information and the attribute reliability, the regularization strength estimation means for estimating the regularization strength of the pixels in the input image and the reconstructed image that is the reconstructed input image using the regularization strength.
- Image reconstructing means for generating a composition image.
- a data processing method calculates an amount of change between a value of a predetermined pixel of an input image and values of pixels around the pixel in an input image to be processed, and is designated in the input image.
- the attribute reliability that is the reliability of the attribute of the pixel of the input image is calculated based on the attribute that is the property of the pixel in the selected area and the change amount, and the image quality information that is the image quality information in the attribute and the attribute trust
- the regularization strength of the pixels in the input image is estimated, and a reconstructed image that is an image obtained by reconstructing the input image using the regularization strength is generated.
- the recording medium includes a process for calculating a change amount between a value of a predetermined pixel of the input image and values of pixels around the pixel in the input image to be processed, and designation in the input image Processing for calculating attribute reliability, which is the reliability of the attribute of the pixel of the input image, based on the attribute which is the property of the pixel in the region and the amount of change, and image quality information which is image quality information in the attribute Based on the attribute reliability, the computer performs processing for estimating the regularization strength of the pixels in the input image and processing for generating a reconstructed image that is an image reconstructed from the input image using the regularization strength.
- FIG. 1 is a block diagram showing an example of the configuration of the information processing apparatus according to the first embodiment of the present invention.
- FIG. 2 is a flowchart illustrating an example of the operation of the information processing apparatus according to the first embodiment.
- FIG. 3 is a block diagram illustrating an example of another configuration of the information processing apparatus according to the first embodiment.
- FIG. 4 is a block diagram illustrating an example of still another configuration of the information processing apparatus according to the first embodiment.
- FIG. 5 is a diagram illustrating an example of an image displayed by the information processing apparatus according to the first embodiment.
- FIG. 6 is a diagram illustrating an example of designation of a teacher information area in the information processing apparatus according to the first embodiment.
- FIG. 7 is a diagram illustrating an example of attribute reliability calculated by the information processing apparatus according to the first embodiment.
- FIG. 8 is a diagram illustrating an example of an image reconstructed by the information processing apparatus according to the first embodiment.
- FIG. 9 is a block diagram illustrating an example of the configuration of the information processing apparatus according to the second embodiment.
- FIG. 10 is a flowchart illustrating an example of the operation of the information processing apparatus according to the second embodiment.
- FIG. 11 is a block diagram illustrating an example of a configuration of the teacher information receiving unit according to the first embodiment.
- the “input image” is an image received by the information processing apparatus according to the embodiment of the present invention.
- Reconstructed image is an image reconstructed (restored) based on the input image.
- the reconstructed image is not limited to an image that is finally output by the information processing apparatus according to the embodiment, but includes an image that is generated as a candidate in the middle of processing.
- an image generated as a candidate may be a final image, the reconstructed image may be referred to as an output image in the following description.
- Degraded image is an image generated by simulating (applying) a degradation process to a reconstructed image.
- the deterioration process is not particularly limited.
- the deterioration process is, for example, blurring (bluring), adding noise (superimposing), or reducing the resolution.
- FIG. 1 is a block diagram showing an example of the configuration of the information processing apparatus 500 according to the first embodiment of the present invention.
- the direction of the arrow in the drawing shows an example, and does not limit the direction of the signal between the blocks.
- the information processing apparatus 500 includes an image processing unit 20, an image receiving unit 10, and an image output unit 30.
- the image receiving unit 10 receives an input image from another device (not shown) (for example, an imaging device such as a camera or a scanner, or a device that processes an image such as an editing device).
- another device for example, an imaging device such as a camera or a scanner, or a device that processes an image such as an editing device.
- the editing device displays a plurality of images on the touch panel. The user operates the touch panel to select an image. Then, the image receiving unit 10 may receive an image selected by the user from the editing device.
- the image receiving unit 10 may record the received input image in a storage unit (for example, a memory) (not shown).
- a storage unit for example, a memory
- the format of the image data in the present embodiment is not particularly limited. However, in the following description, for convenience of explanation, it is assumed that an image is represented using a vertical vector in which pixel values of pixels included in the image are arranged in raster scan order.
- the vertical vector [Y] of the input image received by the image receiving unit 10 is referred to as the input image vector [Y].
- the input image vector [Y] (Y 1 ,..., Y i ,..., Y M ).
- t the superscript “t” described in parentheses means a transposed matrix.
- a vertical vector with M elements is equivalent to an M ⁇ 1 matrix. Therefore, the superscript “t” in the above-mentioned vector means a vertical vector (a vector in which vector elements are arranged vertically).
- the pixel values of the pixels of the reconstructed image are represented using vertical vectors arranged in raster scan order.
- the vertical vector [X] of the reconstructed image is referred to as a reconstructed image vector [X].
- the reconstructed image vector [X] (X 1 ,..., X i ,..., X M ′ ) t .
- the image output unit 30 outputs the reconstructed image reconstructed by the image processing unit 20.
- the image output unit 30 may include, for example, a display device (not shown) and display (output) the reconstructed image on the display device.
- the image output unit 30 may transmit (output) the reconstructed image via a network to a display device (not shown) (for example, the editing device described above).
- the image processing unit 20 generates (reconstructs) a reconstructed image based on the input image. Therefore, the image processing unit 20 includes a teacher information receiving unit 210, a change amount calculating unit 220, an attribute reliability calculating unit 230, a regularization strength estimating unit 240, and an image reconstruction unit 250.
- the image processing unit 20 may include a storage unit (not shown). In that case, each configuration may store each piece of information described later in the storage unit, and take out each piece of information from the storage unit. In the following description, for convenience of explanation, each configuration will be described as passing information directly. However, this does not exclude delivery of information using the storage unit.
- the image processing unit 20 may use a storage unit for delivering some or all of the information.
- the image output unit 30 may acquire a reconstructed image from the storage unit.
- the storage unit may have the same configuration as the storage unit in the description of the image receiving unit 10 described above.
- the teacher information receiving unit 210 In receiving the input image, the teacher information receiving unit 210, together with the input image, the designated area (area information), the attribute in the area (attribute information), and information on the image quality desired in the attribute (image quality information) Receive information including Hereinafter, information including attribute information and image quality information is referred to as “teacher information”. However, the teacher information may include other information (for example, region information or preprocessing information).
- the preprocessing information is information related to processing (preprocessing) executed by the image reconstruction unit 250 described later before image reconstruction.
- the preprocessing is, for example, correction processing (brightness correction, white balance processing, or color correction), filter processing, or noise removal processing.
- “Area information” is information indicating a range designated in the input image.
- the area information in the present embodiment is not particularly limited.
- the area information may be, for example, information representing a predetermined shape in the input image (for example, the center position and radius if circular).
- the area information may be a set of coordinates of the outline of the area.
- the attribute is the property of the image in the specified area.
- the attributes in the present embodiment are not particularly limited.
- the attribute may be a property of a change in pixel value in a region, such as a texture region (a region in which the pixel value repeats regular changes to some extent) or a flat region (a region in which there are few pixel changes).
- the attribute may be the type of an object (object) included in the image, such as a car or a person.
- the attribute may be the type of material of the object included in the image, such as iron or wood.
- the attribute may be a property related to a physical quantity of an object included in an image, such as temperature or density.
- the attribute may be an optical property of an object included in the image, such as brightness or color.
- the attribute may be a property related to the number of objects included in the image, such as the number of people or the number.
- Image quality information is information for setting the regularization strength (that is, the degree of constraint), which will be described in detail later, in the area where the attribute is specified.
- a texture area and a flat area are specified as attribute information given by the user.
- the regularization strength of the texture region is “0” and the regularization strength of the flat region is “1”.
- the regularization strength in this case is a value indicating the degree of suppression (value range is 0 to 1).
- strength is large is an area
- the image quality information may directly set the regularization intensity in the pixel.
- the image quality information is not limited to this.
- the image quality information may be a parameter value in a predetermined calculation formula used for calculating the regularization intensity.
- the teacher information receiving method is not particularly limited.
- the teacher information receiving unit 210 may receive the input teacher information (region information, attribute information, and image quality information) from the editing apparatus in accordance with reception of the input image. Therefore, FIG. 1 shows the connection from the image receiving unit 10 to the teacher information receiving unit 210 by using a broken arrow.
- the teacher information receiving unit 210 may receive information included in the teacher information (region information, attribute information, and image quality information) at a time or may be received individually.
- the operation of the teacher information receiving unit 210 will be described more specifically.
- FIG. 11 is a block diagram illustrating an example of the configuration of the teacher information receiving unit 210.
- the direction of the arrow in the drawing shows an example, and does not limit the direction of the signal between the blocks.
- the teacher information receiving unit 210 includes an area specifying unit 211 and a teacher information acquiring unit 212.
- the area designation unit 211 receives designation of an area for acquiring teacher information in the reconstructed image.
- the area specifying unit 211 includes a display device (for example, a liquid crystal display) and an input device (for example, a touch panel) not shown. Then, the area specifying unit 211 displays the reconstructed image on the display device (liquid crystal display).
- FIG. 5 is a diagram illustrating an example of an image displayed by the area specifying unit 211.
- the user of the information processing apparatus 500 operates the input device (touch panel) of the area specifying unit 211 displaying the image to specify the area (area information) used as the teacher information on the reconstructed image.
- FIG. 6 is a diagram showing an example of designation of an area used as teacher information.
- black lines shown in the eyes, clothing portions, and the like are designations of regions with large changes (for example, texture regions).
- the gray line shown on the cheek around the mouth and the background wall is the designation of an area with little change (for example, a flat area).
- the area specifying unit 211 is not limited to a touch panel as an input device, and other devices such as a mouse or a tablet may be used.
- the area specifying unit 211 transmits the specified area (area information) to the teacher information acquiring unit 212.
- the teacher information acquisition unit 212 acquires attributes (attribute information) in the designated area.
- the teacher information acquisition unit 212 may receive the attribute information for each of the designated areas shown in FIG.
- the area designation unit 211 when the area designation unit 211 receives designation of an area, it may be received including attributes.
- the area designating unit 211 can select a color for designating the area. For example, designation using black is designated as a texture area, and designation using gray is designated as a flat area.
- the area designating unit 211 may receive an input from the touch panel designating the color as the area designation. In that case, the area specifying unit 211 may transmit area information and attribute information regarding the specified area to the teacher information acquiring unit 212.
- the teacher information acquisition unit 212 acquires image quality information (for example, the regularization strength of the pixels of the reconstructed image) regarding the designated area. And the teacher information acquisition part 212 should just process the information received from the area
- image quality information for example, the regularization strength of the pixels of the reconstructed image
- the teacher information receiving unit 210 uses the regularization strength in the input image as the image quality information.
- the present embodiment is not limited to this.
- the teacher information receiving unit 210 may acquire the image quality information as described below.
- the image reconstruction unit 250 described later reconstructs (generates) a reconstructed image based on a plurality of regularization strengths. Then, the teacher information receiving unit 210 transmits a plurality of reconstructed images to a device operated by the user. The user's device displays the received plurality of reconstructed images. The user device may be included in the information processing device 500.
- the device operated by the user requests the user to select an image close to the desired image quality in the case of the texture region and an image close to the desired image quality in the case of the flat region. Then, the device operated by the user notifies the teacher information receiving unit 210 of information (for example, an identifier or number of the image) of each image selected by the user.
- the teacher information receiving unit 210 sets the regularization strength in the image selected as the texture region as the regularization strength when the attribute is the texture region.
- the teacher information receiving unit 210 sets the regularization strength in the image selected as the flat region as the regularization strength when the attribute is the flat region.
- the teacher information receiving part 210 should just use said regularization intensity
- the change amount calculation unit 220 calculates the change amount of the pixel value between the value of the pixel in the input image and the value of the peripheral pixel of the pixel for each pixel.
- the peripheral pixels are pixels adjacent to the target pixel (for example, four pixels including upper and lower, right and left pixels, or eight pixels including diagonally positioned pixels).
- the peripheral pixels may include a predetermined range of pixels adjacent to the adjacent pixels. It is desirable that the change amount calculation unit 220 calculates the change amount for all the pixels of the input image.
- the input image may include an image that is not suitable for processing. In such a case, the change amount calculation unit 220 may calculate the change amount for some pixels of the input image.
- the information processing apparatus 500 receives an instruction for a non-target region from, for example, a device operated by the user. Then, the change amount calculation unit 220 does not have to calculate the change amount of the region.
- the change amount calculated by the change amount calculation unit 220 is not particularly limited.
- the change amount calculation unit 220 may calculate a change vector (magnitude of change) and a direction unit vector (direction of change) as the amount of change as follows.
- the magnitude of change between the i-th pixel and the surrounding pixels is dY i .
- the M ⁇ 2 matrix in which the two direction unit vectors [N] i are vertically arranged in the raster scan order in this way is referred to as a direction unit vector [N].
- the change vector [dY] may be determined so that the filter used for image processing has a larger value as the absolute value of the pixel value of the image obtained by multiply-adding the input image is larger.
- a filter used for image processing is, for example, a Sobel filter, a Prewitt filter, a Laplacian filter, or a Gabor filter.
- a Sobel filter for example, a Sobel filter, a Prewitt filter, a Laplacian filter, or a Gabor filter.
- Two vectors obtained by multiply-adding Sobel filters in the x direction and the y direction on the input image are an x direction change vector [Y x ] and a y direction change vector [Y y ], respectively.
- the y-direction change vector [Y y ] is [Y y1 ,..., Y yi ,..., Y yM ) t .
- the change amount calculation unit 220 applies Y xi and Y yi to Equation 1 shown below to obtain the magnitude of change in the pixel value at the pixel i (i-th element of the displacement vector [dY]). dY i can be calculated.
- the change amount calculation unit 220 calculates the direction unit vector [N] based on the already calculated magnitudes of changes in pixel values in the x and y directions. For example, the change amount calculation unit 220 applies the direction change vector value (Y xi , Y yi ) of the i-th pixel and the magnitude of change (dY i ) to the pixel value in Equations 2 and 3, and i
- the direction unit vector [N] i (N xi , N yi ) of the change in the pixel value of the th pixel is calculated.
- the change amount calculation unit 220 uses the Sobel filter to calculate the edge segment and the magnitude and direction of change in the pixel value on a scale larger than the edge segment has been described.
- the present embodiment is not limited to this.
- the change amount calculation unit 220 may calculate the magnitude and direction of the change in pixel value at a plurality of different scales.
- the change amount calculation unit 220 calculates the magnitude and direction of change in pixel values on L different scales.
- the scale is expressed using a distance between pixels.
- this embodiment is not limited to the distance between pixels as a scale.
- the smallest scale is R 1
- the lth scale from the smallest is R 1 .
- the change amount calculation unit 220 interpolates and enlarges the Sobel filters in the x direction and the y direction by R 1 times.
- the change amount calculation unit 220 generates an image obtained by multiply-adding the Sobel filter expanded by interpolation on the input image.
- the x-direction change vector and the y-direction change vector at this time are [y xl ] and [y yl ], respectively.
- the x-direction change vector [y xl ] is [y xl1 ,..., Y xli ,..., Y xlM ) t It is.
- the change amount calculation unit 220 the size dy li of change in pixel value in the pixel i, with y xli and y yli, calculated as Equation 4 below.
- the change amount calculation unit 220 may generate a vertical vector [dy l ] in which the change magnitudes dy li in each pixel are arranged in the raster scan order as the change magnitudes in the scale.
- the change amount calculation unit 220 may use mathematical expressions similar to the mathematical expressions 2 and 3 for this calculation.
- the method of calculating the magnitude and direction of the change in pixel value at L different scales is not limited to the above description.
- the change amount calculation unit 220 indicates the magnitude and direction of change in the pixel value of the l-th smallest scale in the i-th pixel for each pixel whose distance from the i-th pixel is equal to or less than R l .
- the average value of the magnitude and direction of the change in the pixel value may be used.
- the change amount calculation unit 220 may include the value of the pixel as the magnitude of the change in addition to the magnitude of the change from the surrounding pixels. This is because even if the change has the same magnitude, the effect of the change differs between a small pixel value and a large pixel value.
- the attribute reliability calculation unit 230 calculates the reliability for each attribute of the pixel of the input image using the attribute included in the teacher information and the calculated amount of change in the pixel of the input image. For example, when a texture area and a flat area are designated as attributes, the attribute reliability calculation unit 230 calculates a reliability as a texture area and a reliability as a flat area in the input image. It is desirable that the attribute reliability calculation unit 230 calculates the reliability for all pixels of the input image. However, the attribute reliability calculation unit 230 may calculate the reliability for some of the pixels of the input image, similar to the change amount calculation unit 220 described above.
- the attribute reliability calculation unit 230 calculates the attribute reliability for one or more attributes.
- the method for calculating the reliability in the present embodiment is not particularly limited.
- the method for calculating the reliability may be selected according to the actual desired image.
- the attribute reliability calculation unit 230 constructs a discriminator based on the attribute information and the amount of change in the pixel value corresponding to the attribute information. Then, the attribute reliability calculation unit 230 may calculate the attribute reliability by applying the discriminator to an image portion of an area not included in the pixel area for which the attribute is specified. Note that the attribute reliability calculation unit 230 may determine the pixel region for which the attribute is specified based on the region information.
- the attribute reliability calculation unit 230 may calculate the reliability of the pixel portion for which the attribute is specified. That is, the attribute reliability calculation unit 230 may calculate the reliability for all pixels of the input image without distinguishing the areas. In this case, the attribute reliability calculation unit 230 may not receive the region information.
- the attribute reliability calculation unit 230 may operate as follows.
- an SVM Serial Vector Machine
- Each variable input attribute information e.g., 0 in texture region, such as taking the 1 in a flat region
- t i representing the a xi] i
- the kernel is represented by K (•, •).
- the kernel is a kernel function (inner product in the feature space) in the pattern recognition technique.
- the attribute reliability calculation unit 230 can calculate the attribute reliability ⁇ j of the pixel j by substituting the change amount ⁇ j in the pixel j into Equation 5 shown below.
- Equation 5 the variable i is a subscript for distinguishing elements of the support vector set (S).
- ⁇ in Equation 5 represents the sum of all the elements of the support vector set (S).
- FIG. 7 is a diagram showing an example of attribute reliability calculated based on the teacher information shown in FIG. FIG. 7 shows the reliability of each pixel using brightness.
- the white portion is the portion with the highest reliability as the flat region.
- the black part is the part with the lowest reliability as the flat region.
- a black part is a part with the highest reliability as a texture area
- the regularization strength estimation unit 240 estimates (calculates) the regularization strength ( ⁇ ) in the pixel of the input image based on the attribute reliability calculated by the attribute reliability calculation unit 230 and the image quality information.
- the regularization strength estimation unit 240 desirably estimates the regularization strength ⁇ for all pixels of the input image.
- the attribute reliability calculation unit 230 may estimate the regularization strength ⁇ for some pixels of the input image, similar to the change amount calculation unit 220 described above.
- the method for estimating the regularization strength ⁇ is not particularly limited.
- the attribute information includes a flat region and a texture region
- the image quality information is a regularization strength in the flat region and the texture region.
- the present embodiment is not limited to this.
- the regularization strength estimation unit 240 may calculate the regularization strength ⁇ j at the pixel j using Formula 6 shown below.
- regularized intensity estimating unit 240 Using Equation 6, regularized intensity estimating unit 240, a regularization strength lambda j at pixel j, a value close attribute reliability and high in regularization strength lambda 1 as a predetermined attribute, the predetermined attribute of the If the attribute reliability is low, a value close to ⁇ 0 is calculated.
- a vector in which the regularization intensity ⁇ j in all pixels is arranged in the raster scan order is [ ⁇ ].
- the image quality information is information in which the regularization strength ⁇ 1 of the texture region is “0” and the regularization strength ⁇ 0 of the flat region is “1” as in the example described above.
- the teacher information receiving unit 210 has received image quality information from the user of the information processing apparatus 500 that does not suppress the texture area but suppresses it in the flat area.
- the attribute reliability calculation unit 230 calculates the attribute reliability ⁇ j as the texture region as the reliability ⁇ j for the attribute.
- the information processing apparatus 500 does not need to receive image quality information for all attributes as teacher information.
- the attributes are two types of areas (texture area and flat area).
- the attribute reliability calculation unit 230 may calculate the reliability for the texture region as the attribute reliability ⁇ .
- the regularization strength estimation unit 240 has a low attribute reliability, that is, the regularization strength of a pixel having an attribute that the user does not want (for example, a flat region) is opposite to the specified image quality information.
- the regularization strength estimation unit 240 of the information processing apparatus 500 can estimate (calculate) the regularization strength that provides the user-desired image quality for the pixel of the user-desired attribute based on the teacher information. . Furthermore, the regularization strength estimation unit 240 can estimate (calculate) the regularization strength that leaves a user-desired image quality for pixels having attributes that are not desired by the user.
- the image reconstruction unit 250 generates a reconstructed image that is an image obtained by reconstructing the input image based on the regularization intensity estimated by the regularization intensity estimation unit 240.
- the method for reconstructing an image in the present embodiment is not particularly limited.
- the image reconstruction unit 250 may reconstruct an image using the method described in Patent Document 1.
- the image reconstruction unit 250 may generate a reconstructed image using image enhancement that enhances a specific frequency component in the above-described region where the regularization strength is high as the image reconstruction processing.
- the image reconstruction unit 250 reconstructs (generates) an image using image enhancement processing (for example, high dynamic range imaging) that increases contrast in an area where the regularization intensity is high as image reconstruction processing. )
- the image reconstruction unit 250 determines a regularization term “E reg ([X])” corresponding to the regularization strength applied to each pixel based on the regularization strength at each pixel of the input image.
- the determined regularization term E reg ([X]) may include a component in the direction of change of the pixel value.
- the image reconstruction unit 250 calculates the sum of the determined regularization term E reg ([X]) and the error term E data ([X]) for the input image as shown in Equation 7.
- a certain optimization function (E (X)) is determined. In Equation 7, [] indicating a vector is omitted.
- Equation 7 the error term E data ([X]) is an image (hereinafter referred to as blur) that simulates a deterioration process with respect to the reconstructed image [X] obtained by reconstructing the input image [Y]. This is a function that takes a smaller value as the difference between the input image and the input image is smaller.
- ⁇ in Equation 7 is a parameter determined in advance by the user of the information processing apparatus 500.
- E data ([X]) represents the relationship between the input image [Y] and the reconstructed image [X] using the input image [Y] and the blur matrix [B].
- an image becomes a blurred image, that is, an unclear image due to various factors at the time of photographing the image (the lens of the optical system is out of focus or camera shake).
- the blur function is a function that represents such an effect that an image becomes unclear.
- An example of the blur function is a point spread function (PSF: Point Spread Function).
- PSF Point Spread Function
- the blur function is not limited to the point spread function.
- the blur function may be another function as long as it is a function representing the degree of blur in the blur image.
- the blur function is set in the information processing apparatus 500 in advance as a function determined by the user of the information processing apparatus 500.
- the blur function (or the blur expressed using the blur function) is expressed using an N ⁇ N blur matrix [B].
- N is the number of pixels. That is, the blur matrix [B] is a square matrix including the number of pixels in rows and columns.
- the vertical vector of an image with M pixels is represented by [Z]
- the vertical vector of an image blurred using a given blur function is represented by [Z] b .
- the relationship between the vertical vector [Z] and the vertical vector [Z] b is expressed as Equation 8 below using the blur matrix [B].
- the vertical vector [Z] represents a non-blurred image
- the vertical vector [Z] b represents a blurred image
- the input image [Y] is generally a blurred image. Therefore, the input image [Y] corresponds to the vertical vector [Z] b .
- the reconstructed image [X] corresponds to the vertical vector [Z].
- the error term E data ([X]) is a function including the input image [Y], the reconstructed image [X], and the blur matrix [B].
- the error term E data ([X]) is a function that takes a smaller value as the error between the image (degraded image) blurred by applying the blur function [B] to the reconstructed image [X] and the input image is smaller. is there.
- the error function E data ([X]) can be defined, for example, as a relational expression such as Expression 9 shown below.
- Equation 9 p is a parameter set in advance by the user of the information processing apparatus 500.
- ) on the right side of Equation 9 represents the norm of the vector.
- the norm is a generalized length in analysis.
- the subscript value (p) of the norm (double line) indicates the dimension. That is, “
- the superscript value (p) represents a power.
- Equation 10 the regularization term E reg ([X]) is expressed as, for example, Equation 10 below.
- the matrix [D] is a matrix representing an image differential filter.
- the matrix diag [ ⁇ ] is a diagonal matrix in which the regularization intensity at each pixel calculated by the regularization intensity estimation unit 240 is arranged diagonally.
- the image reconstruction unit 250 generates (searches) a reconstructed image [X] that minimizes the value of the optimization function E ([X]) expressed by Equation 7.
- the search method in the image reconstruction unit 250 is not particularly limited. Examples of the search method include a gradient method or a conjugate gradient method.
- the image reconstruction unit 250 can search for each pixel value of the reconstructed image using these methods.
- the teacher information includes the preprocessing information already described
- the image reconstruction unit 250 executes image preprocessing based on the preprocessing information before image reconstruction using Equation 7. Also good.
- FIG. 8 is a diagram showing an example of an image reconstructed based on the image shown in FIG. Compared with the image shown in FIG. 5, the image shown in FIG. 8 maintains the flatness of the flat areas such as the cheeks and the walls, and the texture areas such as the area around the eyes and the pattern of the clothes. The resolution has improved.
- FIG. 2 is a flowchart showing an example of the operation of the information processing apparatus 500 according to the present embodiment.
- the image receiving unit 10 receives an input image that is an image to be reconstructed (target image) (step S200).
- the change amount calculation unit 220 calculates the change amount in the input image (step S202).
- the attribute reliability calculation unit 230 calculates the attribute reliability based on the change amount and the teacher information (step S203).
- the regularization strength estimation unit 240 estimates the regularization strength based on the attribute reliability (step S204).
- the image reconstruction unit 250 generates (reconstructs) a reconstructed image from the input image using the regularization strength (step S205).
- the image output unit 30 outputs a reconstructed image (step S206).
- the output reconstructed image is displayed on a user device (not shown), for example.
- the user device receives a correction request from the user, the user device transmits the correction request to the information processing device 500.
- the information processing apparatus 500 When the information processing apparatus 500 receives an image correction request from the user (Yes in step S207), the information processing apparatus 500 returns to step S201 and repeats the above processing.
- the information processing apparatus 500 receives at least teacher information as a correction request.
- the teacher information received here may be a part of teacher information (for example, a part of attribute information and image quality information).
- the information processing apparatus 500 ends the process.
- the information processing apparatus 500 according to the first embodiment can achieve an effect that an appropriate parameter in image processing can be adjusted (designated) more easily.
- the change amount calculation unit 220 calculates the change amount of the input image. Then, the attribute reliability calculation unit 230 calculates the attribute reliability based on the teacher information and the change amount. Then, the regularization strength estimation unit 240 estimates the regularization strength based on the teacher information and the attribute reliability. Then, the image reconstruction unit 250 reconstructs the input image using the regularization strength to generate a reconstructed image. As described above, information necessary as an input by the information processing apparatus 500 is teacher information.
- the teacher information may include attribute information (attribute information) of a partial area of the input image and image quality information (image quality information) in the attribute.
- attribute information and the image quality information are not a specification of fine parameter values, but a selection of a certain level of image property and a specification of selection of a desired image quality. That is, the designation of teacher information is a considerably simpler designation than the designation of parameters in a general image. In this way, the information processing apparatus 500 is for reconstructing an image by estimating an appropriate parameter (for example, regularization strength) based on a user's simple designation.
- the user of the information processing apparatus 500 can designate correction of the re-editing process based on the reconstructed image generated by the information processing apparatus 500.
- the information processing apparatus 500 can provide a simpler method in comparison with the parameter setting, that is, image discrimination in the user's image processing setting (instruction) instruction.
- each component of the information processing apparatus 500 may be configured with a hardware circuit.
- each component may be configured using a plurality of devices connected via a network.
- FIG. 3 is a block diagram illustrating an example of the configuration of the information processing apparatus 501 according to the first modification of the present embodiment.
- the direction of the arrow in the drawing shows an example, and does not limit the direction of the signal between the blocks.
- the information processing apparatus 501 includes a change amount calculation unit 220, an attribute reliability calculation unit 230, a regularization strength estimation unit 240, and an image reconstruction unit 250.
- the information processing apparatus 501 receives the input image and teacher information (attribute information and image quality information) via a network (not shown) and the like, operates in the same manner as the image processing unit 20 of the information processing apparatus 500, and reconstructed images Is transmitted to another device via a network (not shown).
- the information processing apparatus 501 may operate in the same manner as the information processing apparatus 500 by reading an input image and teacher information stored in a storage unit (not shown).
- FIG. 3 shows region information with parentheses.
- the information processing apparatus 501 configured in this way can obtain the same effects as the information processing apparatus 500.
- each configuration of the information processing device 501 generates a reconstructed image based on the input image and the teacher information received via the network, similarly to the configuration of the information processing device 500. It is.
- the information processing apparatus 501 is the minimum configuration in the embodiment of the present invention.
- the plurality of components may be configured with one piece of hardware.
- the information processing apparatus 500 may be realized as a computer apparatus including a CPU (Central Processing Unit), a ROM (Read Only Memory), and a RAM (Random Access Memory).
- the information processing apparatus 500 may be realized as a computer apparatus that further includes an input / output connection circuit (IOC: Input / Output Circuit) and a network interface circuit (NIC: Network Interface Circuit).
- IOC Input / Output Circuit
- NIC Network Interface Circuit
- FIG. 4 is a block diagram showing an example of the configuration of the information processing apparatus 600 according to this modification.
- the information processing apparatus 600 includes a CPU 610, a ROM 620, a RAM 630, an internal storage device 640, an IOC 650, and a NIC 680, and constitutes a computer device.
- the CPU 610 reads a program from ROM 620.
- the CPU 610 controls the RAM 630, the internal storage device 640, the IOC 650, and the NIC 680 based on the read program.
- the computer including the CPU 610 controls these configurations, and realizes the function as the image processing unit 20 shown in FIG. That is, the computer including the CPU 610 realizes the functions of the teacher information receiving unit 210, the change amount calculating unit 220, the attribute reliability calculating unit 230, the regularization strength estimating unit 240, and the image reconstruction unit 250. .
- the computer including the CPU 610 may further realize functions as the image receiving unit 10 and the image output unit 30 illustrated in FIG. 1.
- the CPU 610 may use the RAM 630 or the internal storage device 640 as a temporary storage of a program when realizing each function.
- the CPU 610 may read a program included in the recording medium 700 storing the program so as to be readable by a computer by using a recording medium reading device (not shown).
- the CPU 610 may receive a program from an external device (not shown) via the NIC 680, store the program in the RAM 630, and operate based on the stored program.
- ROM 620 stores programs executed by CPU 610 and fixed data.
- the ROM 620 is, for example, a P-ROM (Programmable-ROM) or a flash ROM.
- the RAM 630 temporarily stores programs executed by the CPU 610 and data.
- the RAM 630 is, for example, a D-RAM (Dynamic-RAM).
- the internal storage device 640 stores data and programs stored in the information processing device 600 for a long period of time. Further, the internal storage device 640 may operate as a temporary storage device for the CPU 610.
- the internal storage device 640 is, for example, a hard disk device, a magneto-optical disk device, an SSD (Solid State Drive), or a disk array device.
- the IOC 650 mediates data between the CPU 610, the input device 660, and the display device 670.
- the IOC 650 is, for example, an IO interface card or a USB (Universal Serial Bus) card.
- the input device 660 is a device that receives an input instruction from an operator of the information processing apparatus 600.
- the input device 660 is, for example, a keyboard, a mouse, or a touch panel.
- the input device 660 may function as the image receiving unit 10 or the teacher information receiving unit 210.
- the display device 670 is a device that displays information to the operator of the information processing apparatus 600.
- the display device 670 is a liquid crystal display, for example.
- the display device 670 may operate as the image output unit 30 or the teacher information receiving unit 210 (region specifying unit 211).
- the NIC 680 relays data exchange with an external device (not shown) via the network.
- the NIC 680 is, for example, a LAN (Local Area Network) card.
- the information processing apparatus 600 configured in this way can obtain the same effects as the information processing apparatus 500.
- FIG. 9 is a block diagram illustrating an example of the configuration of the information processing apparatus 510 according to the second embodiment.
- the direction of the arrow in the drawing shows an example, and does not limit the direction of the signal between the blocks.
- the information processing apparatus 510 according to the second embodiment is compared with the information processing apparatus 500 according to the first embodiment, and the image processing unit 21 is replaced with the image processing unit 20. It differs in that it includes.
- the image processing unit 21 includes a change amount calculation unit 221 and an attribute reliability calculation unit 231 instead of the change amount calculation unit 220 and the attribute reliability calculation unit 230, and further includes a learning image.
- the difference is that a receiving unit 261 and a learning image teacher information receiving unit 271 are included. Therefore, the description of the same configuration and operation as in the first embodiment will be omitted, and the configuration and operation unique to this embodiment will be described.
- the information processing apparatus 510 may be realized using a computer shown in FIG.
- the learning image receiving unit 261 receives one or more images (learning images) different from the input image.
- the learning image teacher information receiving unit 271 receives the teacher information corresponding to the learning image in the same manner as the teacher information receiving unit 210 receives the teacher information (first teacher information) corresponding to the input image. That is, the learning image teacher information receiving unit 271 receives teacher information (second teacher information) including attribute information of some pixels of the learning image and image quality information desired in the attribute.
- the attribute information included in the second teacher information may be the same attribute information as the attribute information included in the first teacher information, or may be different attribute information.
- the attribute information included in the second teacher information includes various attributes described in the first embodiment already described (for example, pixel properties, object types, materials, physical quantities, optical properties). Or information on numbers). Further, the second teacher information may include area information or preprocessing information.
- the image reconstruction unit 250 may perform preprocessing using preprocessing information included in the second teacher information.
- the learning image receiving unit 261 and the learning image teacher information receiving unit 271 may use all of the received learning image and teacher information, or a part of them.
- the learning image receiving unit 261 and the learning image teacher information receiving unit 271 calculate the similarity between the input image and the learning image, and correspond to the learning image having the similarity in a predetermined range and the learning image.
- Teacher information (second teacher information) may be used.
- the calculation method of the similarity between the input image and the learning image used by the learning image receiving unit 261 and the learning image teacher information receiving unit 271 is not particularly limited.
- the learning image reception unit 261 and the learning image teacher information reception unit 271 use the image feature amount (shift feature amount, Fisher vector, etc.) used in general image processing to determine the similarity. It may be calculated.
- the change amount calculation unit 221 calculates the change amount based on the input image and the learning image.
- the change amount calculation unit 221 may apply the same method as the change amount calculation unit 220 of the first embodiment to the input image and the learning image as a method of calculating the change amount.
- the attribute reliability calculation unit 231 calculates the reliability corresponding to each attribute in all the pixels of the input image using the attribute in the input image and the learning image and the change amount calculated by the change amount calculation unit 221.
- FIG. 10 is a flowchart showing an example of the operation of the information processing apparatus 510.
- the same operations as those in FIG. 2 are denoted by the same reference numerals, and detailed description thereof is omitted.
- the learning image receiving unit 261 receives the learning image
- the learning image teacher information receiving unit 271 receives teacher information corresponding to the learning image (step S400).
- the image receiving unit 10 receives an input image (step S200), and the teacher information receiving unit 210 receives teacher information for the input image (step S201).
- the change amount calculation unit 221 calculates the change amount based on the input image and the learning image (step S402).
- the attribute reliability calculation unit 231 calculates the attribute reliability based on the change amount and the teacher information of the input image and the learning image (S402).
- the information processing apparatus 510 according to the second embodiment can achieve an effect that more accurate processing can be realized.
- the learning image receiving unit 261 receives the learning image. Further, the learning image teacher information receiving unit 271 receives teacher information corresponding to the learning image.
- the change amount calculation unit 221 calculates the change amount based on the learning image in addition to the input image. That is, the change amount calculation unit 221 calculates the change amount based on more images than the change amount calculation unit 220. Therefore, the change amount calculation unit 221 can calculate a change amount with higher accuracy than the change amount calculation unit 220.
- the attribute reliability calculation unit 231 is based on the teacher information corresponding to the learning image and the amount of change with high accuracy described above. Attribute reliability is calculated. Therefore, the attribute reliability calculation unit 231 can calculate attribute reliability with higher accuracy than the attribute reliability calculation unit 230.
- the regularization strength estimation unit 240 estimates the regularization strength based on the attribute reliability with high accuracy, it can estimate the regularization strength with higher accuracy.
- the image reconstruction unit 250 can generate a reconstructed image with higher accuracy because it uses the regularization strength with high accuracy.
- the learning image receiving unit 261 receives an image such as a monitoring camera image, a medical image, or a satellite image as a learning image.
- the learning image teacher information receiving unit 271 receives, as teacher information, for example, an attribute transmitted by the user of the information processing apparatus 510 using the user's apparatus and image quality information desired in the attribute. .
- the learning image teacher information receiving unit 271 may receive, as teacher information, an attribute and image quality information desired in the attribute from a device operated by a person different from the user.
- the medical worker who has more experience in analyzing medical images than the medical worker can provide learning information (attribute information and image quality information) that is optimal for analyzing the medical image.
- the information may be transmitted to the information processing apparatus 510.
- the attribute information is information indicating whether the pixel is an affected part, for example.
- the image quality information is, for example, information in which the regularization intensity of the region that is not the affected area is “1”.
- the learning image teacher information receiving unit 271 receives the teacher information including the preprocessing information, the medical staff can share the experience and know-how regarding the preprocessing.
- the learning image and the teacher information (attribute information and image quality information) received by the learning image receiving unit 261 and the learning image teacher information receiving unit 271 are other devices (for example, a server) Or the like may be stored in a recording device).
- the information processing apparatus 510 may use information stored in the apparatus based on an instruction from a user of the information processing apparatus 510 as necessary.
- the information processing apparatus 510 may store the change amount calculated by the change amount calculation unit 221 instead of the learning image received by the learning image reception unit 261.
- recording the amount of change instead of the learning image enables users to share experiences and know-how regarding mutual image analysis while maintaining confidentiality of the image.
- the present invention can be applied to the use of analyzing criminal investigation images or satellite images.
- the present invention can also be applied to uses for analyzing medical images.
- the present invention can also be applied to applications in which image analysis experience or know-how of image analysts is shared based on the regularization strength after calculation or use of teacher information in each image analysis.
Abstract
Description
次に、発明における第1の実施の形態について図面を参照して説明する。
図1は、本発明における第1の実施の形態に係る情報処理装置500の構成の一例を示すブロック図である。図面中の矢印の方向は、一例を示すものであり、ブロック間の信号の向きを限定するものではない。
次に、図面を参照して本実施の形態に係る情報処理装置500の動作について説明する。
次に、本実施の形態の効果について説明する。
以上の説明した情報処理装置500は、次のように構成される。
また、情報処理装置500において、複数の構成部は、1つのハードウェアで構成されてもよい。
次に、図面を参照して、第2の実施の形態について説明する。
図9は、第2の実施の形態に係る情報処理装置510の構成の一例を示すブロック図である。図面中の矢印の方向は、一例を示すものであり、ブロック間の信号の向きを限定するものではない。
次に、図面を参照して本実施の形態に係る情報処理装置510の動作について説明する。
次に、第2の実施の形態の効果について説明する。
次に、第二の実施の形態に係る情報処理装置510のより具体的な動作例について説明する。
20 画像処理部
21 画像処理部
30 画像出力部
210 教師情報受信部
211 領域指定部
212 教師情報取得部
220 変化量算出部
221 変化量算出部
230 属性信頼度算出部
231 属性信頼度算出部
240 正則化強度推定部
250 画像再構成部
261 学習用画像受信部
271 学習画像用教師情報受信部
500 情報処理装置
501 情報処理装置
510 情報処理装置
600 情報処理装置
610 CPU
620 ROM
630 RAM
640 内部記憶装置
650 IOC
660 入力機器
670 表示機器
680 NIC
700 記録媒体
Claims (10)
- 処理の対象である入力画像において、前記入力画像の所定の画素の値と、前記画素の周辺の画素の値との変化量を算出する変化量算出手段と、
前記入力画像における指定された領域の画素の性質である属性と、前記変化量とを基に、前記入力画像の画素の属性についての信頼度である属性信頼度を算出する属性信頼度算出手段と、
前記属性における画質の情報である画質情報と、前記属性信頼度とを基に、前記入力画像における画素の正則化強度を推定する正則化強度推定手段と
前記正則化強度を用いて前記入力画像を再構成した画像である再構成画像を生成する画像再構成手段と
を含む情報処理装置。 - 前記画像再構成手段が、
前記正則化強度に基づく正則化の強さを画素毎に表す正則化項と、前記再構成画像を劣化させた画像である劣化画像と前記入力画像との画素の差分を表す誤差項とに基づいて、前記再構成画像を生成する
請求項1に記載の情報処理装置。 - 前記正則化強度推定手段が、
各画素において、前記属性信頼度が高いほど前記属性における前記画質情報に沿うように前記正則化強度を推定し、前記属性信頼度が低いほど前記属性における前記画質情報から離れるような前記正則化強度を推定する
請求項1又は2に記載の情報処理装置。 - 前記属性信頼度算出手段が、
前記属性が指定された領域の情報である領域情報を基に、
前記属性の信頼度を算出する
請求項1ないし3のいずれか1項に記載の情報処理装置。 - 前記入力画像を受信する画像受信手段と、
前記再構成画像を出力する画像出力手段と、
前記属性の情報と前記画質情報とを含む教師情報を受信する教師情報受信手段とをさらに含み、
前記教師情報受信手段が、
前記画像出力手段が出力した前記再構成画像における所定の領域の画素の前記正則化強度を前記画質情報として受信する
請求項1ないし4のいずれか1項に記載の情報処理装置。 - 前記画像再構成手段が、複数の再構成画像を生成し、
前記画像出力手段が、前記複数の再構成画像を出力し、
前記教師情報受信手段が、前記画像出力手段が出力した複数の前記再構成画像における、複数の領域の画素の前記正則化強度を前記画質情報として受信する
請求項5に記載の情報処理装置。 - 前記教師情報受信手段が、
前記再構成画像における前記画素の正則化強度の領域を指定する領域指定手段と、
前記指定された領域の教師情報として前記領域の画素の属性と正則化強度とを取得する教師情報取得手段とをさらに含む、
請求項5又は6に記載の情報処理装置。 - 前記入力画像とは異なる画像である学習用画像を受信する学習用画像受信手段と、
前記学習用画像に対応する第2の教師情報を受信する学習画像用教師情報受信手段とをさらに含み、
前記変化量算出手段が、
前記学習用画像を基に前記変化量を算出し、
前記属性信頼度算出手段が、
前記学習用画像の第2の教師情報を用いて前記属性信頼度を算出する
請求項1ないし7のいずれか1項に記載の情報処理装置。 - 処理の対象である入力画像において、前記入力画像の所定の画素の値と、前記画素の周辺の画素の値との変化量を算出し、
前記入力画像における指定された領域の画素の性質である属性と、前記変化量とを基に、前記入力画像の画素の属性についての信頼度である属性信頼度を算出し、
前記属性における画質の情報である画質情報と、前記属性信頼度とを基に、前記入力画像における画素の正則化強度を推定し、
前記正則化強度を用いて前記入力画像を再構成した画像である再構成画像を生成する
情報処理方法。 - 処理の対象である入力画像において、前記入力画像の所定の画素の値と、前記画素の周辺の画素の値との変化量を算出する処理と、
前記入力画像における指定された領域の画素の性質である属性と、前記変化量とを基に、前記入力画像の画素の属性についての信頼度である属性信頼度を算出する処理と、
前記属性における画質の情報である画質情報と、前記属性信頼度とを基に、前記入力画像における画素の正則化強度を推定する処理と、
前記正則化強度を用いて前記入力画像を再構成した画像である再構成画像を生成する処理と
をコンピュータに実行させるプログラムをコンピュータに読み取り可能に記録する記録媒体。
Priority Applications (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
JP2016564678A JP6645442B2 (ja) | 2014-12-17 | 2015-12-10 | 情報処理装置、情報処理方法、及び、プログラム |
US15/529,547 US10540735B2 (en) | 2014-12-17 | 2015-12-10 | Information processing device, information processing method, and recording medium |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
JP2014255158 | 2014-12-17 | ||
JP2014-255158 | 2014-12-17 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2016098323A1 true WO2016098323A1 (ja) | 2016-06-23 |
Family
ID=56126231
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/JP2015/006170 WO2016098323A1 (ja) | 2014-12-17 | 2015-12-10 | 情報処理装置、情報処理方法、及び、記録媒体 |
Country Status (3)
Country | Link |
---|---|
US (1) | US10540735B2 (ja) |
JP (1) | JP6645442B2 (ja) |
WO (1) | WO2016098323A1 (ja) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109564675A (zh) * | 2016-07-25 | 2019-04-02 | 日本电气株式会社 | 信息处理设备、信息处理方法以及记录介质 |
KR20200026435A (ko) * | 2018-09-03 | 2020-03-11 | 네이버 주식회사 | 스킴―픽셀cnn을 이용한 이미지 자동 생성 방법 및 이미지 자동 생성 장치 |
Families Citing this family (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR20200127766A (ko) * | 2019-05-03 | 2020-11-11 | 삼성전자주식회사 | 영상 처리 장치 및 그 영상 처리 방법 |
CN113706425A (zh) * | 2021-06-02 | 2021-11-26 | 上海深至信息科技有限公司 | 一种图像增强系统及便携式多普勒成像仪 |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2010072901A (ja) * | 2008-09-18 | 2010-04-02 | Toshiba Corp | 画像処理装置およびその方法 |
WO2013011797A1 (ja) * | 2011-07-21 | 2013-01-24 | 日本電気株式会社 | 劣化復元システム、劣化復元方法およびプログラム |
-
2015
- 2015-12-10 US US15/529,547 patent/US10540735B2/en active Active
- 2015-12-10 WO PCT/JP2015/006170 patent/WO2016098323A1/ja active Application Filing
- 2015-12-10 JP JP2016564678A patent/JP6645442B2/ja active Active
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2010072901A (ja) * | 2008-09-18 | 2010-04-02 | Toshiba Corp | 画像処理装置およびその方法 |
WO2013011797A1 (ja) * | 2011-07-21 | 2013-01-24 | 日本電気株式会社 | 劣化復元システム、劣化復元方法およびプログラム |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109564675A (zh) * | 2016-07-25 | 2019-04-02 | 日本电气株式会社 | 信息处理设备、信息处理方法以及记录介质 |
CN109564675B (zh) * | 2016-07-25 | 2023-12-19 | 日本电气株式会社 | 信息处理设备、信息处理方法以及记录介质 |
KR20200026435A (ko) * | 2018-09-03 | 2020-03-11 | 네이버 주식회사 | 스킴―픽셀cnn을 이용한 이미지 자동 생성 방법 및 이미지 자동 생성 장치 |
KR102103727B1 (ko) | 2018-09-03 | 2020-04-24 | 네이버 주식회사 | 스킴―픽셀cnn을 이용한 이미지 자동 생성 방법 및 이미지 자동 생성 장치 |
Also Published As
Publication number | Publication date |
---|---|
US20170316542A1 (en) | 2017-11-02 |
JPWO2016098323A1 (ja) | 2017-10-05 |
JP6645442B2 (ja) | 2020-02-14 |
US10540735B2 (en) | 2020-01-21 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Liu et al. | Robust color guided depth map restoration | |
US20090110285A1 (en) | Apparatus and method for improving image resolution using fuzzy motion estimation | |
JP6645442B2 (ja) | 情報処理装置、情報処理方法、及び、プログラム | |
US8903195B2 (en) | Specification of an area where a relationship of pixels between images becomes inappropriate | |
JP6390173B2 (ja) | 情報処理装置、情報処理システム、画像処理方法、及び、プログラム | |
Tian et al. | Joint image registration and super-resolution from low-resolution images with zooming motion | |
US10475229B2 (en) | Information processing apparatus and information processing method | |
US20200118250A1 (en) | Image processing apparatus, image processing method, and non-transitory computer-readable storage medium | |
Jeong et al. | Multi-frame example-based super-resolution using locally directional self-similarity | |
Wang et al. | Blurred image restoration using knife-edge function and optimal window Wiener filtering | |
JPWO2018225133A1 (ja) | 画像処理装置、画像処理方法、および画像処理プログラム | |
CN113256529A (zh) | 图像处理方法、装置、计算机设备及存储介质 | |
JP7328096B2 (ja) | 画像処理装置、画像処理方法、およびプログラム | |
JP6581359B2 (ja) | 情報処理装置及びその制御方法及びプログラム及び記憶媒体、並びに、ビデオカメラ | |
WO2019123554A1 (ja) | 画像処理装置、画像処理方法、及び、記録媒体 | |
KR101341617B1 (ko) | 단일 영상의 오차모델을 기반으로 한 고해상도 영상 복원장치 및 방법 | |
JP4887491B2 (ja) | 医用画像処理方法及びその装置、プログラム | |
EP3932319A1 (en) | Learning method, learning device, generative model and program | |
JP2017010094A (ja) | 画像処理装置、撮像装置、画像処理方法、画像処理プログラム、および、記憶媒体 | |
JP6997969B2 (ja) | 画像合成システム、画像合成方法、および画像合成プログラム記録媒体 | |
Khattab et al. | Regularized multiframe Super-Resolution image reconstruction using linear and nonlinear filters | |
US20130120461A1 (en) | Image processor and image processing method | |
Wang et al. | Fast convergence strategy for multi-image superresolution via adaptive line search | |
JP2023003763A (ja) | 学習装置、画像処理装置、学習処理方法、及びプログラム | |
Gedkhaw et al. | Superresolution Reconstruction in Automatic Thai Sign Language Feature Extraction Using Adaptive Triangulation Interpolation. |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 15869532 Country of ref document: EP Kind code of ref document: A1 |
|
ENP | Entry into the national phase |
Ref document number: 2016564678 Country of ref document: JP Kind code of ref document: A |
|
WWE | Wipo information: entry into national phase |
Ref document number: 15529547 Country of ref document: US |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 15869532 Country of ref document: EP Kind code of ref document: A1 |