WO2015115168A1 - 画像処理装置 - Google Patents
画像処理装置 Download PDFInfo
- Publication number
- WO2015115168A1 WO2015115168A1 PCT/JP2015/050663 JP2015050663W WO2015115168A1 WO 2015115168 A1 WO2015115168 A1 WO 2015115168A1 JP 2015050663 W JP2015050663 W JP 2015050663W WO 2015115168 A1 WO2015115168 A1 WO 2015115168A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- signal
- value
- similarity
- image
- resolution pattern
- Prior art date
Links
- 238000012545 processing Methods 0.000 title claims abstract description 255
- 238000000034 method Methods 0.000 claims abstract description 132
- 238000004364 calculation method Methods 0.000 claims abstract description 51
- 238000001914 filtration Methods 0.000 abstract 1
- 238000012986 modification Methods 0.000 description 44
- 230000004048 modification Effects 0.000 description 44
- 238000003672 processing method Methods 0.000 description 14
- 238000010586 diagram Methods 0.000 description 12
- 230000002708 enhancing effect Effects 0.000 description 6
- 238000009499 grossing Methods 0.000 description 4
- 238000007796 conventional method Methods 0.000 description 3
- 238000004590 computer program Methods 0.000 description 2
- 238000013500 data storage Methods 0.000 description 2
- 238000003707 image sharpening Methods 0.000 description 2
- 239000000047 product Substances 0.000 description 2
- 238000013459 approach Methods 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
- 239000013589 supplement Substances 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T3/00—Geometric image transformations in the plane of the image
- G06T3/40—Scaling of whole images or parts thereof, e.g. expanding or contracting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/20—Image enhancement or restoration using local operators
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/73—Deblurring; Sharpening
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N5/00—Details of television systems
- H04N5/14—Picture signal circuitry for video frequency region
- H04N5/20—Circuitry for controlling amplitude response
- H04N5/205—Circuitry for controlling amplitude response for correcting amplitude versus frequency characteristic
- H04N5/208—Circuitry for controlling amplitude response for correcting amplitude versus frequency characteristic for compensating for attenuation of high frequency components, e.g. crispening, aperture distortion correction
-
- 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
- G06T2207/20024—Filtering details
Definitions
- the present invention relates to an image processing technique, for example, a technique for appropriately improving a detail portion in an image on which an upscaling process has been executed.
- a low resolution image (video) such as SD (Standard Definition) image quality is displayed on a high resolution display that displays a high resolution image (video) such as HD (High Definition) image quality.
- a high resolution image (video) such as HD (High Definition) image quality.
- an up-scaling process for extending (enlarging) the input image (video) is performed.
- an image (video) that has just undergone upscaling is displayed on a high-resolution display, it is displayed as a blurred image (video) compared to a true HD image (video).
- a process for sharpening the image is performed on the image (video) after the upscaling process. There is.
- Patent Document 1 Japanese Patent No. 5281690
- a high frequency component is extracted from an input image signal, a limit process is performed on the extracted high frequency component, and a high frequency after the limit process is performed.
- the component is added to the input image signal.
- Patent Document 1 only high-frequency components included in the original signal (input image signal) are emphasized and processed, so that image sharpening processing is appropriately performed. You may not be able to.
- a thin line portion such as hair becomes thick.
- image signal even if the high frequency component included in the image (image signal) is emphasized as in the prior art, the thickened line remains thick. It is only emphasized (for example, it is emphasized so that the outline is clear), and it is difficult to reproduce a thin line before upscaling.
- the high frequency component included in the original signal is only emphasized and processed, so the signal included in the original signal (input image signal). It is difficult to reproduce a line that has been thickened by processing for emphasizing a component beyond the band or an upscaling process into an original thin line. Further, in the conventional technique, a process of adding a high frequency component included in the image signal to the input image signal is executed, so noise is unnecessarily emphasized with respect to the input image signal having a large amplitude. Also, there is a problem in that overshoot occurs excessively for an input image signal having a large amplitude.
- An object of the present invention is to realize an image processing apparatus capable of reproducing a linear image area into an original thin linear image area.
- a first configuration is an image processing device that sharpens an input image, and includes an amplitude estimation unit, a similarity calculation unit, a multiplication unit, an FIR filter unit, and an addition unit. .
- An amplitude estimation unit downscales the high resolution pattern data string, and then adjusts the low resolution pattern data string obtained by multiplying the low resolution pattern data string obtained by the upscaling process by a magnification factor, and The magnification a when the difference from the pixel data string in the local area, which is a predetermined range of the input image including the processing target pixel, is equal to or less than a predetermined value is acquired.
- the similarity calculation unit calculates the similarity value of the processing target pixel as the adjusted low resolution pattern data string and the pixel data string as the difference between the adjusted low resolution pattern data string and the pixel data string is smaller in the local region. Is set to a value indicating that the degree of similarity is high.
- the multiplication unit multiplies the magnification a and the similarity for each processing target.
- the FIR filter unit uses the difference pattern data sequence acquired by subtracting the high resolution pattern data sequence from the low resolution pattern data sequence as an FIR filter coefficient, and is similar to the magnification a acquired by the multiplication unit for each processing target pixel. By executing the FIR filter process on the multiplication value with the degree, the high-frequency component data ⁇ D of the processing target pixel is acquired.
- the addition unit adds, for each processing target pixel, the pixel data of the processing target pixel and the high frequency component data ⁇ D of the processing target pixel acquired by the FIR filter processing.
- a linear image region that has become thicker as a result of upscaling a low-resolution image without unnecessarily enhancing noise components or causing excessive overshoot is possible to realize an image processing apparatus that can reproduce the original thin linear image region.
- FIG. 1 is a schematic configuration diagram of an image processing system 1000 according to a first embodiment. The figure for demonstrating the principle of a thinning process.
- the flowchart of the image processing method of 2nd Embodiment. The flowchart of the image processing method of 2nd Embodiment.
- FIG. 1 is a schematic configuration diagram of an image processing system 1000 according to the first embodiment.
- the image processing system 1000 includes an image processing unit (image processing apparatus) 100 and a pattern generation unit 200 as shown in FIG.
- the image processing unit 100 includes an amplitude estimation unit 1, a similarity calculation unit 2, a multiplication unit 3, an FIR filter unit 4, and an addition unit 5.
- the amplitude estimation unit 1 inputs an input image (input image signal) Din and a low resolution pattern signal PtnL output from the pattern generation unit 200.
- the amplitude estimating unit 1 acquires the amplitude estimated value a based on the input image signal Din and the low resolution pattern signal PtnL, and the acquired estimated amplitude value a (signal indicating the estimated amplitude value a) to the multiplying unit 3. Output.
- the similarity calculation unit 2 inputs an input image (input image signal) Din and a low resolution pattern signal PtnL output from the pattern generation unit 200.
- the similarity calculation unit 2 acquires the similarity value s based on the input image signal Din and the low resolution pattern signal PtnL, and uses the acquired similarity value s (signal indicating the similarity value s) as the multiplication unit 3. Output to.
- the multiplication unit 3 inputs the estimated amplitude value a (estimated amplitude value signal a) output from the amplitude estimation unit 1 and the similarity value s (similarity value signal s) output from the similarity calculation unit 2. .
- the multiplication unit multiplies the input estimated amplitude value a (estimated amplitude value signal a) and the similarity value s (similarity value signal s), and outputs the signal acquired by the multiplication process to the FIR filter unit 4. .
- the FIR filter unit 4 includes the difference pattern signal PtnDiff output from the pattern generation unit 200, the estimated amplitude value a (estimated amplitude value signal a) and the similarity value s (similarity value signal s) output from the multiplication unit 3. ) And a multiplication result signal.
- the FIR filter unit 4 sets the filter coefficient of the FIR filter based on the difference pattern signal PtnDiff, and executes FIR filter processing on the signal (multiplication result signal) output from the multiplication unit 3. Then, the FIR filter unit 4 outputs the signal after the FIR filter processing to the adding unit 5 as a difference signal ⁇ D.
- the addition unit 5 inputs the input image (input image signal) Din and the difference signal ⁇ D output from the FIR filter unit 4. Then, the adding unit 5 adds the input image signal Din and the difference signal ⁇ D, and outputs the signal acquired by the addition process as the image signal Dout.
- the pattern generation unit 200 includes a downscale processing unit 201, an upscale processing unit 202, and a subtraction unit 23, as shown in FIG.
- the downscale processing unit 201 receives the high resolution pattern signal PtnH and executes a downscale process on the input high resolution pattern signal PtnH. Then, the downscale processing unit 201 outputs the signal after the downscale processing to the upscale processing unit 202.
- the upscaling processing unit 202 receives the signal output from the downscaling processing unit 201, and performs upscaling processing on the input signal. Then, the upscale processing unit 202 outputs the signal after the up processing to the amplitude estimation unit 1 and the similarity calculation unit 2 of the image processing unit 100 as the low resolution pattern signal PtnL. The upscale processing unit 202 also outputs the low resolution pattern signal PtnL to the subtracting unit 23.
- the subtraction unit 23 receives the high resolution pattern signal PtnH and the low resolution pattern signal PtnL output from the upscale processing unit 202, and performs a process of subtracting the high resolution pattern signal PtnH from the low resolution pattern signal PtnL. Then, the subtraction unit 23 outputs the signal acquired by the subtraction process to the FIR filter unit 4 of the image processing unit 100 as the difference pattern signal PtnDiff.
- a high-resolution image for example, HD video
- an image (video) that has been upscaled again is taken as an example. This will be described with reference to FIG.
- FIG. 2 is a diagram for explaining the principle of the thinning process.
- data on a one-dimensional axis (for example, a horizontal axis) of a two-dimensional image is extracted and shown as one-dimensional data for the sake of simplicity.
- the image signals A to D are drawn with the horizontal axis as the coordinate position of the one-dimensional axis (eg, horizontal axis) and the vertical axis as the pixel value (eg, luminance value).
- the image signal A is, for example, a line having a line width w1 (extending in a direction perpendicular to the one-dimensional axis direction) in a one-dimensional axis (for example, horizontal axis) direction (for example, in a vertical axis direction).
- This is an image signal for forming an image corresponding to an extended line.
- Image signal B is an image signal after downscale processing (for example, decimating processing and smoothing processing) is performed on image signal A.
- the image signal C is an image signal after an upscaling process (for example, an interpolating process and a smoothing process) is performed on the image signal B. That is, the image signal C is an image signal after the true HD image quality image signal A is once downscaled to become the image signal B, and the image signal B is upscaled again.
- an upscaling process for example, an interpolating process and a smoothing process
- the portion corresponding to the line having the line width w1 has a peak pixel value (luminance value) of about half in the image signal C, and the line width is w2 Thickness is about three times thicker.
- the width (line width) w2 of the line portion of the image signal C can be restored to the fineness of the line width w1 of the image signal A that is the original image, a sense of fineness is provided in the image formed by the image signal C. Obtainable.
- the signal indicating the pattern of the image signal C is the low resolution pattern signal PtnL, and a signal pattern (signal series) similar to the low resolution pattern signal PtnL appears in the input image signal Din.
- the signal indicating the pattern of the image signal D is added to the input image signal Din.
- the image processing system 1000 acquires a fine image (video) such as the image signal A.
- the pattern generation unit 200 converts the high resolution pattern signal PtnH corresponding to the image signal A in FIG. 2 to the low resolution pattern signal PtnL corresponding to the image signal C in FIG. A difference pattern signal PtnDiff corresponding to the second image signal D (difference signal D) is generated. Specifically, the pattern generation unit 200 executes processing as follows.
- the high resolution pattern signal PtnH is input to the downscale processing unit 201 and the subtraction unit 23.
- the high resolution pattern signal PtnH is subjected to downscale processing (for example, decimating processing and smoothing processing). Then, the signal that has undergone the downscaling process is output to the upscaling processing unit 202.
- downscale processing for example, decimating processing and smoothing processing.
- the upscale processing unit 202 obtains the low resolution pattern signal PtnL by performing upscaling processing (for example, interpolating processing and smoothing processing) on the signal output from the downscaling processing unit 201. Is done.
- the acquired low resolution pattern signal PtnL is output to the subtraction unit 23, the amplitude estimation unit 1 of the image processing unit 100, and the similarity calculation unit 2 of the image processing unit 100.
- the difference pattern signal PtnDiff is acquired by a process corresponding to.
- the difference pattern signal PtnDiff acquired by the subtracting unit 23 is output to the FIR filter unit 4 of the image processing unit 100.
- the amplitude estimation process is executed on the input image signal Din using the low resolution pattern signal PtnL.
- the amplitude estimation unit 1 calculates the magnification a (amplitude a) that minimizes the difference between the input image signal Din and the signal obtained by multiplying the low resolution pattern signal PtnL by a and b. Is calculated. This will be described with reference to FIG.
- the input image signal Din is f (x)
- the low resolution pattern signal PtnL is k (u)
- the amplitude estimation unit 1 Find the amplitude a (x) and the offset b (x) that minimizes.
- the amplitude a (x) is a real number and a negative value is allowed.
- the sigma symbol in the above equation means that sigma calculation (addition processing) is performed at N sample points in the local region u (the same applies hereinafter).
- the amplitude estimation unit 1 calculates the amplitude a (x) and the offset b (x) that minimize the square error E according to the above equation by the method of least squares. Specifically, the amplitude estimation unit 1 Thus, the amplitude a (x) and the offset b (x) that minimize the square error E are calculated.
- S k , S k2 , A 1, A 2, B 1, and B 2 are constants when the low resolution pattern signal k (u) is determined, and therefore may be set in advance. preferable.
- the calculation processing may be executed by hardware processing, for example, with the configuration shown in FIG. 4 is a delay device (for example, a D flip-flop, a delay device (line memory) for one line, or the like).
- the minimum square error Emin will be described.
- the extreme value condition when the square error E takes the minimum value is that the partial differential value of the square error E with respect to a is “0”, and the partial differential value of the square error E with respect to b is “ 0 ”. That is, the extreme value conditions when the square error E takes the minimum value are as follows. From (Equation 5) to (Equation 7) above, It becomes.
- the above calculation process may be executed by a hardware process, for example, with the configuration shown in FIGS. 5 and 6.
- the similarity calculation process is performed on the input image signal Din shown in FIG. 7, using the low resolution pattern signal PtnL (for example, Ptn1 to Ptn5 shown in FIG. 7) as shown in FIG. Verify similarity.
- AR (x) indicates a local region whose center is x, and in the case of FIG. 8, the number of pixels (number of samples) included in the local region is “9”.
- the FIR filter processing is executed using the difference pattern signal PtnDiff.
- the FIR filter unit 4 executes FIR filter processing using the data string of the difference pattern signal PtnDiff as the FIR filter coefficient. That is, when the differential pattern signal PtnDiff is d (x + i) (i: integer, ⁇ n ⁇ i ⁇ n), the FIR filter unit 4 The signal ⁇ D (x) after the FIR filter processing is acquired by executing the processing corresponding to.
- the FIR filter processing may be executed by hardware processing with the configuration shown in FIG. 9, for example.
- the summing unit 41 in FIG. 9 performs addition of each input (sigma calculation).
- (X)) signal waveform of the signal a (x) k (u) + b (x) (a signal obtained by setting the amplitude of the low resolution pattern signal to a (x) and adding the offset value) in FIG. 10, “PtnL ⁇ a (x1) ”, (3) the signal waveform of the output signal ⁇ D ( ⁇ D (x)) of the FIR filter unit 4, and (5) the signal waveform of the output signal Dout.
- (X coordinate axes) are shown to coincide. Further, the low resolution pattern signal PtnL and the difference pattern signal PtnDiff are shown on the left side of FIG.
- the error between the signal waveform when the amplitude estimation unit 1 multiplies the low resolution pattern signal PtnL by a and adds the offset b and the input image signal Din is the most.
- the magnification a is set so as to decrease.
- the similarity calculator 2 causes the least square error between the signal waveform when the low resolution pattern signal PtnL is multiplied by a and the offset b is added in the local region and the input image signal.
- the multiplication signal s ⁇ a generated by the multiplication unit 3 is input to the FIR filter unit 4, so that the FIR filter unit 4 multiplies the signal value of the differential pattern signal PtnDiff by a.
- the signal ⁇ D is output.
- the adder 5 adds the signal ⁇ D obtained by multiplying the signal value of the difference pattern signal PtnDiff by a and the input image signal Din.
- the high resolution pattern signal PtnH is downscaled, and further, similar to the signal waveform using a signal waveform obtained by multiplying the low resolution pattern signal PtnL when the upscale processing is performed.
- the signal waveform portion to be detected is detected in the input image signal.
- a signal waveform obtained by multiplying the difference pattern signal PtnDiff which is a difference between the high resolution pattern signal PtnH and the low resolution pattern signal PtnL, is a in the signal portion of the input image signal detected as described above. to add.
- a signal waveform portion similar to a signal waveform obtained by multiplying the low resolution pattern signal PtnL by a is a portion in which a high frequency component is lost by downscaling and further upscaling.
- the lost high frequency component is similar to a signal waveform (signal ⁇ D) obtained by multiplying the differential pattern signal PtnDiff by a.
- the difference pattern signal PtnDiff is multiplied by a with respect to the input image signal Din (the magnification a is a difference between the signal waveform obtained by multiplying the low resolution pattern signal PtnL by a and the input image signal Din).
- the magnification a is a difference between the signal waveform obtained by multiplying the low resolution pattern signal PtnL by a and the input image signal Din.
- the image processing system 1000 As described above, in the image processing system 1000, a portion having a signal waveform similar to the signal processing waveform generated when the downscaling process is performed and the upscaling process is performed is detected, and only the detected part is lost.
- the high-frequency component is adjusted to be equivalent to the high-frequency component signal whose amplitude is lost, and is added to the input image signal. Therefore, in the image processing system 1000, as in the conventional technology, the low-resolution image is thickened by upscaling without unnecessarily enhancing noise components or causing excessive overshoot. It is possible to reproduce the curled linear image area into the original thin linear image area. As a result, the image signal (video signal) acquired by the image processing system 1000 becomes a high-quality image signal (video signal) with a fine feeling.
- the image processing system according to the present modification has the same configuration as that of the image processing system 1000 according to the first embodiment, and only the processing executed by the amplitude estimation unit 1 is different.
- FIG. 12 is a diagram for explaining processing of the amplitude estimation unit 1 in the present modification.
- the amplitude estimation unit 1 in this modification example simply obtains the amplitude a instead of obtaining the amplitude a by the least square method as in the first embodiment.
- the amplitude estimation unit 1 Thus, the average value AveP of the pixel values (signal values) of the pixels included in the area AR_p and the average value AveF of the pixel values (signal values) of the pixels included in the areas AR_f1 and AR_f2 are calculated.
- n is the number of samples before and after the center of the area AR_p.
- m is the number of samples before and after the center in the region constituted by the region AR_p, the region AR_f1, and the region AR_f2.
- the processing content of the similarity determination unit 21 of the similarity calculation unit 2 of the image processing system 1000 of the first embodiment is different in the image processing system of this modification.
- the similarity determination unit 21 of the first embodiment has the configuration shown in FIG. 6 and determines whether or not the processing target pixel takes the minimum value of Emin in the local region.
- the similarity determination unit 21 of this modification is configured by a ROM or the like having input / output characteristics as shown in FIG. 13 instead of the configuration shown in FIG. That is, as shown in FIG. 13, the similarity determination unit 21 of the present modification has an output value, that is, a similarity of “1” as the value of the input least square error Emin approaches “0”. Get closer and get bigger.
- processing can be performed using the similarity s having an intermediate value between 0 and 1.
- the input / output characteristics of the similarity determination unit 21 of the present modification are not limited to the above, and other characteristics may be used as long as the characteristics monotonously decrease with respect to the value of the input minimum square error Emin. It may be. Further, the similarity determination unit 21 of the present modification may be realized by a ROM or the like, or may realize the above input / output characteristics by calculation.
- the image processing system of this modification has a configuration in which the similarity calculation unit 2 is replaced with a similarity calculation unit 2A shown in FIG. 14 in the image processing system 1000 of the first embodiment. Other than that, the image processing system of this modification is the same as the image processing system 1000 of the first embodiment.
- the processing target pixel takes a peak value (maximum value or minimum value) in the local region of the input image Din (in the case of FIG. 14, a local region including nine pixels).
- FIG. 15 shows the signal waveform of the input image Din and the similarity s acquired by the similarity calculation unit 2A of the present modification, with the x-axis being matched.
- the similarity s is appropriately calculated (determined) by the similarity calculator 2A of the present modification.
- the local region is a local region composed of nine pixels (sample points).
- the similarity s can be easily calculated in the image processing system according to the present modification.
- FIG. 16 shows a schematic configuration diagram of an image processing system 1000A according to a fourth modification of the first embodiment.
- the image processing system 1000A of this modification has a configuration in which the pattern generation unit 200 of the image processing system 1000 of the first embodiment is replaced with a pattern generation unit 200A.
- the image processing system 1000 ⁇ / b> A according to the present modification includes a low-pass filter in which the downscale processing unit 201 and the upscale processing unit 202 are included in the pattern generation unit 200 of the first embodiment.
- the unit 204 is replaced.
- the image processing system 1000A of the present modification is the same as the image processing system 1000 of the first embodiment.
- the low-pass filter unit 204 receives the high-resolution pattern signal PtnH and performs low-pass filter processing (low-pass filter processing) on the input high-resolution pattern signal PtnH. Then, the low-pass filter unit 204 outputs the signal after the low-pass filter process to the subtraction unit 23 and the amplitude estimation unit 1 as a low resolution pattern signal PtnL.
- the high-resolution pattern signal PtnH is downscaled on the signal waveform acquired when the low-pass filter processing is executed and the high-resolution pattern signal PtnH. Thereafter, the low-pass filter so that the signal waveform obtained when the upscaling process is similar (for example, the difference (error) between the two (two signal waveforms) falls within a predetermined range). It is preferable to set a cut-off frequency for processing.
- the high-resolution pattern signal PtnH is acquired by performing the downscaling process by the downscaling processing unit 201 and then performing the upscaling process by the upscaling processing unit 202.
- the low resolution pattern signal PtnL is a signal obtained by reducing the high frequency component of the high resolution pattern signal PtnH. That is, this signal approximates a signal obtained by performing low-pass filter processing on the high resolution pattern signal PtnH.
- the low-pass pattern signal PtnL is obtained by executing the low-pass filter process on the high-resolution pattern signal PtnH by the low-pass filter unit 204 as described above. . Therefore, the low resolution pattern signal PtnL acquired in the image processing system 1000A of this modification approximates the low resolution pattern signal PtnL acquired in the image processing system 1000 of the first embodiment. As a result, in the image processing system 1000A, processing similar to that in the first embodiment can be realized by performing processing using the low resolution pattern signal PtnL acquired by the pattern generation unit 200A.
- the low-resolution pattern signal PtnL can be acquired only by performing the low-pass filter processing by the low-pass filter unit 204. Compared to the embodiment, the amount of calculation can be reduced.
- the configuration of the image processing system 1000 of the first embodiment may be adopted.
- a high-precision low-resolution pattern signal PtnL is acquired by executing a down-scale process by the down-scale processing unit 201 and an up-scale process by the up-scale processing unit 202 on the high-resolution pattern signal PtnH.
- the configuration of the image processing system 1000A of this modification may be adopted.
- the low-resolution pattern signal PtnL can be obtained only by performing the low-pass filter processing by the low-pass filter unit 204 for the high-resolution pattern signal PtnH, so that the amount of calculation can be reduced and necessary.
- the computer resources that are used can be reduced.
- 17 to 20 are flowcharts of the image processing method according to the second embodiment.
- the resolution estimation method according to the present embodiment is executed by, for example, the image processing system according to the first embodiment, or a system including a control unit (such as a CPU), an image memory, a ROM, and a RAM.
- a control unit such as a CPU
- an image memory such as a ROM
- a RAM such as a RAM
- step S1 In step S1, a pattern (low resolution pattern) k (u) obtained by reducing the resolution of the high resolution pattern is set (calculated).
- step S2 In step S2, a difference pattern d (u) that is a difference between the high resolution pattern and the low resolution pattern is set (calculated).
- step S3 In step S3, A1, A2, B1, and B2 are calculated from the low resolution pattern k (u) in the same manner as described in (Formula 3) of the first embodiment.
- step S4 the output image Y (i, j) is initialized with the input image I (i, j).
- step S5 a horizontal sharpening process (horizontal sharpening process) of the image is executed.
- the horizontal sharpening process is executed by adding the high-frequency component ⁇ Y acquired by the horizontal sharpening process to the image Y (i, j).
- step S6 vertical sharpening processing (vertical sharpening processing) is performed on the image.
- the vertical sharpening process is executed by adding the high-frequency component ⁇ Y acquired by the vertical sharpening process to the image Y (i, j).
- steps S5 and S6 may be changed.
- V_SIZE is the vertical image size
- j is the vertical coordinate position of the processing target pixel.
- H_SIZE is the horizontal direct image size, and i is the horizontal coordinate position of the processing target pixel.
- i and j have the above meanings in the following flowcharts.
- step S504 the local area information is stored in f ( ⁇ n: n) (for example, f ( ⁇ n: n) is a memory area in which ⁇ n to n data (2n + 1 data) can be stored). make a copy.
- f ( ⁇ n: n) is a memory area in which ⁇ n to n data (2n + 1 data) can be stored.
- step S505 pixels adjacent in the vertical direction centering on the pixel (i, j) are copied to f ( ⁇ n: n).
- This process f ([ ⁇ n: n]) I (i [ ⁇ n: n], j) Is written.
- step S506 Sfk and Sf are obtained based on the mathematical formula shown in the first embodiment, and the amplitude a is further obtained.
- step S507 the offset value b and Sf2 are obtained based on the mathematical formula shown in the first embodiment, and the least square error Emin is obtained in the same manner as in the first embodiment.
- step S508 the data of amplitude a is stored in the data storage address of amplitude a corresponding to (i, j) in the image memory.
- This process Ma (i, j) a Is written.
- step S521 it is determined whether it is horizontal processing.
- step S524 the temporary variable tmp is cleared, and in step S525, the process is repeated while incrementing the local reference variable u by +1 from -n to + n.
- step S525 the difference pattern d (u) and as (u) are multiplied, and the value of the multiplication result is cumulatively added to tmp. This corresponds to a so-called FIR filter process.
- the sharpening process (horizontal sharpening process, vertical sharpening process) in steps S5 and S6 is executed.
- the image processing method according to the present embodiment a portion having a signal waveform similar to the signal processing waveform generated when the downscaling process and the upscaling process are performed is detected. Only the lost high frequency component is adjusted to be equivalent to the lost high frequency component signal and added to the input image signal. Therefore, in the image processing method of the present embodiment, unlike the conventional technique, the low-resolution image is upscaled without unnecessarily enhancing noise components or causing excessive overshoot. The thickened linear image area can be reproduced as the original thin linear image area. As a result, the image signal (video signal) acquired by the image processing method of the present embodiment is a high-quality image signal (video signal) with a fine feeling.
- the present invention is not limited to this, and some or all of the high resolution pattern signal PtnH, the low resolution pattern signal PtnL, and the differential pattern signal PtnDiff may be set in advance. In this case, for example, in the image processing system, the pattern generation unit 200 may be omitted.
- the pattern (signal waveform) of a part or all of the high resolution pattern signal PtnH, the low resolution pattern signal PtnL, and the difference pattern signal PtnDiff may be changed by setting.
- the image processing system has been described on the assumption that the processing in the horizontal direction is performed.
- vertical processing may be executed.
- an image processing system an image processing apparatus, and an image processing method may be realized by combining a part or all of the above-described embodiments (including modifications).
- part or all of the image processing system and the image processing apparatus according to the above-described embodiment may be realized as an integrated circuit (for example, an LSI, a system LSI, or the like).
- Part or all of the processing of each functional block in the above embodiment may be realized by a program. And a part or all of the processing of each functional block of the above embodiment may be executed by a central processing unit (CPU) in the computer.
- a program for performing each processing is stored in a storage device such as a hard disk or ROM, and a central processing unit (CPU) reads the program from the ROM or RAM and executes it. Also good.
- each process of the above embodiment may be realized by hardware, or may be realized by software (including a case where it is realized together with an OS (operating system), middleware, or a predetermined library). Further, it may be realized by mixed processing of software and hardware.
- OS operating system
- middleware middleware
- predetermined library predetermined library
- execution order of the processing methods in the above embodiment is not necessarily limited to the description of the above embodiment, and the execution order can be changed without departing from the gist of the invention.
- a computer program that causes a computer to execute the above-described method and a computer-readable recording medium that records the program are included in the scope of the present invention.
- the computer-readable recording medium include a flexible disk, a hard disk, a CD-ROM, an MO, a DVD, a large-capacity DVD, a next-generation DVD, and a semiconductor memory.
- the computer program is not limited to the one recorded on the recording medium, but may be transmitted via a telecommunication line, a wireless or wired communication line, a network represented by the Internet, or the like.
- the first invention is an image processing apparatus that sharpens an input image, and includes an amplitude estimation unit, a similarity calculation unit, a multiplication unit, an FIR filter unit, and an addition unit.
- the amplitude estimating unit adjusts the low-resolution pattern data sequence obtained by multiplying the low-resolution pattern data sequence obtained by executing a predetermined process on the high-resolution pattern data sequence by a magnification a,
- the magnification a when the difference from the pixel data string in the local area that is a predetermined range of the input image including the target pixel is equal to or smaller than a predetermined value is acquired.
- the similarity calculation unit calculates the similarity value of the processing target pixel as the adjusted low resolution pattern data string and the pixel data string as the difference between the adjusted low resolution pattern data string and the pixel data string is smaller in the local region. Is set to a value indicating that the degree of similarity is high.
- the multiplication unit multiplies the magnification a and the similarity for each processing target.
- the FIR filter unit uses the difference pattern data sequence acquired by subtracting the high resolution pattern data sequence from the low resolution pattern data sequence as an FIR filter coefficient, and is similar to the magnification a acquired by the multiplication unit for each processing target pixel. By executing the FIR filter process on the multiplication value with the degree, the high-frequency component data ⁇ D of the processing target pixel is acquired.
- the addition unit adds, for each processing target pixel, the pixel data of the processing target pixel and the high frequency component data ⁇ D of the processing target pixel acquired by the FIR filter processing.
- the low-resolution image is thickened by upscaling without unnecessarily enhancing noise components or causing excessive overshoot. It is possible to reproduce the curled linear image area into the original thin linear image area. As a result, the image signal (image data) acquired by the image processing apparatus becomes a high-quality image signal (image data) with a fine feeling.
- the “local region that is a predetermined range of the input image including the processing target pixel” includes, for example, (1) 2n + 1 pixels (n: natural number) in the horizontal direction of the image with the processing target pixel at the center.
- An image area, (2) an image area composed of 2n + 1 pixels (n: natural number) in the vertical direction of the image centered on the processing target pixel, and (3) (2n + 1) pixels ⁇ ( 2n + 1) is a concept including an image area composed of pixels.
- the second invention is the first invention, wherein the amplitude estimation unit downscales the high resolution pattern data string and then upscales to obtain the low resolution pattern data string.
- the low resolution pattern data string can be acquired by performing the downscaling process on the high resolution pattern data string and then performing the upscaling process.
- 3rd invention is 1st invention, Comprising: An amplitude estimation part acquires a low-resolution pattern data sequence by performing a low-pass filter process with respect to a high-resolution pattern data sequence.
- the low-resolution pattern data string can be acquired by executing the low-pass filter process, so that the low-resolution pattern data string can be acquired with a small amount of calculation.
- low-pass filter processing refers to a signal waveform (data sequence pattern) acquired when the high-resolution pattern data sequence is executed and the high-resolution pattern data sequence. After the downscaling process, the signal waveform (data string pattern) obtained when the upscaling process is similar (for example, the difference (error) between the two (two signal waveforms) is within a predetermined range. It is preferable that a cutoff frequency of the low-pass filter process is set.
- the fourth invention is any one of the first to third inventions, wherein the similarity calculation unit (1) has a minimum difference between the adjusted low-resolution pattern data string and the pixel data string in the local region.
- the similarity corresponding to the pixel is set to the first value, and (2) a pixel in which the difference between the adjusted low resolution pattern data string and the pixel data string is not minimized within the local region
- the similarity corresponding to the pixel is set to a value smaller than the first value.
- the similarity degree of the pixel that minimizes the difference between the adjusted low-resolution pattern data string and the pixel data string can be set to a large value in the local region.
- the fifth invention is any one of the first to fourth inventions, and the amplitude estimating unit obtains the magnification a by the least square method.
- a sixth invention is any one of the first to fifth inventions, wherein the similarity calculation unit includes the adjusted low-resolution pattern data string adjusted by the magnification a calculated by the least square method, and the local region Is calculated for each pixel in the local area, and (1) the similarity of the pixel having the minimum square error value Emin is the first value in the local area. (2) In the local region, the similarity of pixels whose minimum square error value Emin is not the minimum value is set to a value smaller than the first value.
- the low-resolution pattern data string after adjustment adjusted by the magnification a calculated by the least square method and the least square error value Emin between the pixel data string in the local region are appropriately used.
- the degree of similarity of the processing target pixel can be set.
- the seventh invention is the invention according to any one of the first to sixth inventions, wherein the similarity calculation unit includes a low-resolution pattern data string after adjustment adjusted by the magnification a, and a pixel data string in the local region. As the difference is smaller, the similarity of the processing target pixel is set to a larger value.
- this image processing apparatus it is possible to set a similarity that takes an intermediate value (for example, an intermediate value between 0 and 1).
- the eighth invention is any one of the first to seventh inventions, in which the similarity calculation unit (1) is configured to process the pixel to be processed when the pixel data of the pixel to be processed has a peak value in the local region. (2) When the pixel data of the processing target pixel is not the peak value in the local region, the similarity corresponding to the processing target pixel is set to be higher than the first value. Set to a smaller value.
- a ninth invention is an image processing method for sharpening an input image, Low resolution obtained by executing predetermined processing (for example, (1) downscale processing and then upscale processing or (2) low-pass filter processing) on the high resolution pattern data string
- predetermined processing for example, (1) downscale processing and then upscale processing or (2) low-pass filter processing
- the difference between the adjusted low-resolution pattern data sequence obtained by multiplying the pattern data sequence by the magnification a and the pixel data sequence in the local area that is a predetermined range of the input image including the processing target pixel is a predetermined value.
- a similarity calculation step for setting a value indicating that the degree of similarity of is high A multiplication step of multiplying the magnification a and the similarity for each processing target; Using the difference pattern data sequence acquired by subtracting the high resolution pattern data sequence from the low resolution pattern data sequence as an FIR filter coefficient, the magnification a and the magnification a acquired by the multiplication unit for each pixel to be processed
- a linear image region that has become thick due to the upscaling processing of a low-resolution image without unnecessarily enhancing noise components or causing excessive overshooting is performed.
- An image processing apparatus and an image processing method that can be reproduced in a thin linear image region can be realized. Therefore, the present invention is useful in the video related industry field and can be implemented in this field.
- Image processing system 100 Image processing unit (image processing apparatus) DESCRIPTION OF SYMBOLS 1 Amplitude estimation part 2, 2A similarity calculation part 3 Multiplication part 4 FIR filter part 5 Addition part
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Signal Processing (AREA)
- Data Mining & Analysis (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Image Processing (AREA)
- Television Systems (AREA)
- Picture Signal Circuits (AREA)
Abstract
Description
第1実施形態について、図面を参照しながら、以下、説明する。
図1は、第1実施形態に係る画像処理システム1000の概略構成図である。
以上のように構成された画像処理システム1000の動作について、以下、説明する。
(画像信号D)=(画像信号A)―(画像信号C)
である。
PtnDiff=PtnH―PtnL
に相当する処理により、差分パターン信号PtnDiffを取得する。
画像処理部100の振幅推定部1では、入力画像信号Dinに対して、低解像度パターン信号PtnLを用いて、振幅推定処理が実行される。
画像処理部100の類似度算出部2では、入力画像信号Dinに対して、低解像度パターン信号PtnLを用いて、類似度算出処理が実行される。
(1)局所領域AR(x1)において、x=x1のときEminは、最小値Emin(x1)となり、
(2)局所領域AR(x2)において、x=x2のときEminは、最小値Emin(x2)となり、
(3)局所領域AR(x3)において、x=x3のときEminは、最小値Emin(x3)となり、
(4)局所領域AR(x4)において、x=x4のときEminは、最小値Emin(x4)となる。
画像処理部100のFIRフィルタ部4では、乗算部3から出力される、振幅a(=a(x))と類似度s(=s(x))の乗算信号(a×s)に対して、差分パターン信号PtnDiffを用いて、FIRフィルタ処理が実行される。
(1)x=x1において、乗算部3の出力信号s×a=a(x1)であるので、このとき、FIRフィルタ部4から、差分パターン信号PtnDiffをa(x1)倍した信号(差分信号ΔD(=ΔD(x)))が出力される。
(2)x=x2において、乗算部3の出力信号s×a=a(x2)であるので、このとき、FIRフィルタ部4から、差分パターン信号PtnDiffをa(x2)倍した信号(差分信号ΔD(=ΔD(x)))が出力される。
(3)x=x3において、乗算部3の出力信号s×a=a(x3)であるので、このとき、FIRフィルタ部4から、差分パターン信号PtnDiffをa(x3)倍した信号(差分信号ΔD(=ΔD(x)))が出力される。
(4)x=x4において、乗算部3の出力信号s×a=a(x4)であるので、このとき、FIRフィルタ部4から、差分パターン信号PtnDiffをa(x4)倍した信号(差分信号ΔD(=ΔD(x)))が出力される。
Dout=Din+ΔD
に相当する処理が実行され、出力画像信号Doutが取得される。
次に、第1実施形態の第1変形例について説明する。
a(x0)=(AveP-AveF)/a0
により、処理対象画素x=x0における、振幅a(x)(=a(x0))を算出する。
次に、第1実施形態の第2変形例について説明する。
次に、第1実施形態の第3変形例について説明する。
次に、第1実施形態の第4変形例について説明する。
次に、第2実施形態について、説明する。
ステップS1では、高解像度パターンを低解像度化したパターン(低解像度パターン)k(u)が設定される(算出される)。
ステップS2では、高解像度パターンと低解像度パターンとの差分である差分パターンd(u)が設定(算出)される。
ステップS3では、低解像度パターンk(u)から、第1実施形態の(数式3)で説明したのと同様に、A1、A2、B1、B2が算出される。なお、(数式3)におけるNは、局所領域に含まれるサンプル数を示す。また、N=2×n+1とし、nは、処理対象画素(注目画素)の画素位置xを中心として、局所領域の範囲を示す値である。つまり、局所領域は、(x-n)~(x+n)で表される。
ステップS4では、出力画像Y(i,j)を、入力画像I(i,j)で初期化する。
ステップS5では、画像の水平方向の先鋭化処理(水平先鋭化処理)が実行される。画像Y(i,j)に、水平先鋭化処理により取得された高域成分ΔYが加算されることにより、水平先鋭化処理が実行される。
次に、ステップS5、S6の先鋭化処理(水平先鋭化処理、垂直先鋭化処理)の詳細について、図18~図20を用いて、説明する。
ステップS501では、j=0とし、jがV_SIZE-1となるまで、jを+1ずつインクリメントしながら、処理を繰り返す。なお、V_SIZEは垂直画像サイズであり、jは処理対象画素の垂直座標位置である。
ステップS504では、局所領域の情報をf(-n:n)(例えば、f(-n:n)は、-n~nのデータ(2n+1個のデータ)を保存することができるメモリ領域)にコピーする。水平処理の場合には、画素(i,j)を中心として水平方向に隣接する画素が、f(-n:n)にコピーされる。この処理を、
f([-n:n])=I(i,j+[-n:n])
と表記する。
f([-n:n])=I(i[-n:n],j)
と表記する。
ステップS506では、第1実施形態で示した数式に基づいて、Sfk、Sfを求め、さらに、振幅aを求める。
Ma(i,j)=a
と表記する。
Me(i,j)=Emin
と表記する。
S509では、iを+1インクリメントし、処理をS502に戻し、上記処理(S503~S508までの処理)を繰り返す。
S510では、jを+1インクリメントし、処理をS501に戻し、上記処理(S502~S509までの処理)を繰り返す。
ステップS511では、j=0とし、jがV_SIZE-1となるまで、jを+1ずつインクリメントしながら、処理を繰り返す。
ステップS513では、処理対象画素(i,j)における最小二乗誤差値Me(i,j)が局所領域内で最小値をとるか否かを判定する。判定の結果、局所領域内において、最小二乗誤差値Me(i,j)が最小値である場合、類似度としてs=1をセットし(ステップS514)、そうでない場合は、s=0をセットする(ステップS515)。
Msa(i,j)=s×Ma(i,j)
に相当する処理を実行する。
S517では、iを+1インクリメントし、処理をS512に戻し、上記処理(S513~S516までの処理)を繰り返す。
S518では、jを+1インクリメントし、処理をS511に戻し、上記処理(S512~S517までの処理)を繰り返す。
ステップS511では、j=0とし、jがV_SIZE-1となるまで、jを+1ずつインクリメントしながら、処理を繰り返す。
ステップS521では、水平処理であるか否かを判定する。
as([-n:n])=Mas(i+[-n:n],j)
に相当する処理が実行される(S523)。
as([-n:n])=Mas(i,j+[-n:n])
に相当する処理が実行される(S522)。
ステップS524では、一時変数tmpをクリアし、ステップS525では、局所参照変数uを-nから+nまで、+1ずつインクリメントしながら処理を繰り返す。このループ処理の中では、差分パターンd(u)と、as(u)とが乗算され、乗算結果の値がtmpに累積加算される。これは、いわゆるFIRフィルタの処理に相当する。
Y(i,j)=Y(i,j)+ΔY
に相当する処理が実行される。
S530では、iを+1インクリメントし、処理をS520に戻し、上記処理(S521~S529までの処理)を繰り返す。
S531では、jを+1インクリメントし、処理をS519に戻し、上記処理(S520~S530までの処理)を繰り返す。
上記実施形態(変形例を含む)では、低解像度パターン信号PtnL、および、差分パターン信号PtnDiffが、高解像度パターン信号PtnHから所定の処理が実行されることで、取得される場合について、説明した。しかしながら、これに限定されることはなく、高解像度パターン信号PtnH、低解像度パターン信号PtnL、および、差分パターン信号PtnDiffの一部または全部が、予め設定されているものであってもよい。この場合、例えば、画像処理システムにおいて、パターン生成部200を省略した構成とすることも可能である。
なお、本発明は、以下のようにも表現することができる。
高解像度パターンデータ列に対して、所定の処理(例えば、(1)ダウンスケール処理した後、アップスケール処理を行う処理や、(2)低域通過フィルタ処理)を実行して取得される低解像度パターンデータ列に倍率aを乗算することにより取得された調整後低解像度パターンデータ列と、処理対象画素を含む入力画像の所定の範囲である局所領域内の画素データ列との差が、所定の値以下となるときの前記倍率aを取得する振幅推定ステップと、
前記局所領域内において、前記調整後低解像度パターンデータ列と前記画素データ列との差が少ない程、前記処理対象画素の類似度値を、前記調整後低解像度パターンデータ列と前記画素データ列との類似の程度が高いことを示す値に設定する類似度算出ステップと、
前記処理対象ごとに、前記倍率aと前記類似度とを乗算する乗算ステップと、
前記低解像度パターンデータ列から前記高解像度パターンデータ列を減算することで取得される差分パターンデータ列をFIRフィルタ係数として、前記処理対象画素ごとに、前記乗算部により取得された前記倍率aと前記類似度との乗算値に対して、FIRフィルタ処理を実行することで、前記処理対象画素の高域成分データΔDを取得するFIRフィルタステップと、
前記処理対象画素ごとに、当該処理対象画素の画素データと、前記FIRフィルタ処理により取得された前記処理対象画素の高域成分データΔDとを加算する加算ステップと、
を備える画像処理方法である。
100 画像処理部(画像処理装置)
1 振幅推定部
2、2A 類似度算出部
3 乗算部
4 FIRフィルタ部
5 加算部
Claims (8)
- 入力画像を先鋭化する画像処理装置であって、
高解像度パターンデータ列に対して所定の処理を実行することで取得される低解像度パターンデータ列に倍率aを乗算することにより取得された調整後低解像度パターンデータ列と、処理対象画素を含む入力画像の所定の範囲である局所領域内の画素データ列との差が、所定の値以下となるときの前記倍率aを取得する振幅推定部と、
前記局所領域内において、前記調整後低解像度パターンデータ列と前記画素データ列との差が少ない程、前記処理対象画素の類似度値を、前記調整後低解像度パターンデータ列と前記画素データ列との類似の程度が高いことを示す値に設定する類似度算出部と、
前記処理対象ごとに、前記倍率aと前記類似度とを乗算する乗算部と、
前記低解像度パターンデータ列から前記高解像度パターンデータ列を減算することで取得される差分パターンデータ列をFIRフィルタ係数として、前記処理対象画素ごとに、前記乗算部により取得された前記倍率aと前記類似度との乗算値に対して、FIRフィルタ処理を実行することで、前記処理対象画素の高域成分データΔDを取得するFIRフィルタ部と、
前記処理対象画素ごとに、当該処理対象画素の画素データと、前記FIRフィルタ処理により取得された前記処理対象画素の高域成分データΔDとを加算する加算部と、
を備える画像処理装置。 - 前記振幅推定部は、前記高解像度パターンデータ列をダウンスケール処理した後、アップスケール処理することで、前記低解像度パターンデータ列を取得する、
請求項1に記載の画像処理装置。 - 前記振幅推定部は、前記高解像度パターンデータ列に対して、低域通過フィルタ処理を実行することで、前記低解像度パターンデータ列を取得する、
請求項1に記載の画像処理装置。 - 前記類似度算出部は、
(1)前記局所領域内において、前記調整後低解像度パターンデータ列と前記画素データ列との差が最小となる画素に対して、当該画素に対応する類似度を第1の値に設定し、(2)前記局所領域内において、前記調整後低解像度パターンデータ列と前記画素データ列との差が最小とならない画素に対して、当該画素に対応する類似度を前記第1の値よりも小さい値に設定する、
請求項1から3のいずれかに記載の画像処理装置。 - 振幅推定部は、最小二乗法により、前記倍率aを求める、
請求項1から4のいずれかに記載の画像処理装置。 - 類似度算出部は、最小二乗法により算出した前記倍率aで調整された前記調整後低解像度パターンデータ列と、前記局所領域内の前記画素データ列との最小二乗誤差値Eminを、前記局所領域内の画素ごとに算出し、(1)前記局所領域内において、前記最小二乗誤差値Eminが最小値である画素の類似度を第1の値に設定し、(2)前記局所領域内において、前記最小二乗誤差値Eminが最小値でない画素の類似度を前記第1の値よりも小さい値に設定する、
請求項1から5のいずれかに記載の画像処理装置。 - 類似度算出部は、
前記倍率aで調整された前記調整後低解像度パターンデータ列と、前記局所領域内の前記画素データ列との差が小さい程、前記処理対象画素の類似度を大きな値に設定する、
請求項1から6のいずれかに記載の画像処理装置。 - 類似度算出部は、
(1)前記局所領域内において、前記処理対象画素の画素データがピーク値であるとき、前記処理対象画素に対応する類似度を第1の値に設定し、(2)前記局所領域内において、前記処理対象画素の画素データがピーク値ではないとき、前記処理対象画素に対応する類似度を第1の値よりも小さい値に設定する、
請求項1から7のいずれかに記載の画像処理装置。
Priority Applications (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US15/036,868 US9679355B2 (en) | 2014-01-28 | 2015-01-13 | Image processing device |
JP2015559850A JP6158956B2 (ja) | 2014-01-28 | 2015-01-13 | 画像処理装置 |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
JP2014-013010 | 2014-01-28 | ||
JP2014013010 | 2014-01-28 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2015115168A1 true WO2015115168A1 (ja) | 2015-08-06 |
Family
ID=53756745
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/JP2015/050663 WO2015115168A1 (ja) | 2014-01-28 | 2015-01-13 | 画像処理装置 |
Country Status (3)
Country | Link |
---|---|
US (1) | US9679355B2 (ja) |
JP (1) | JP6158956B2 (ja) |
WO (1) | WO2015115168A1 (ja) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2020241337A1 (ja) * | 2019-05-24 | 2020-12-03 | 株式会社日立製作所 | 画像処理装置 |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112381812A (zh) * | 2020-11-20 | 2021-02-19 | 深圳市优象计算技术有限公司 | 一种简单高效的图像质量评估方法及系统 |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2007074307A (ja) * | 2005-09-06 | 2007-03-22 | Nippon Telegr & Teleph Corp <Ntt> | 時間解像度の拡大方法及び装置及びプログラム及びコンピュータ読み取り可能な記録媒体 |
JP2011188478A (ja) * | 2010-02-09 | 2011-09-22 | Panasonic Corp | 超解像処理装置及び超解像処理方法 |
JP2013021635A (ja) * | 2011-07-14 | 2013-01-31 | Sony Corp | 画像処理装置、画像処理方法、プログラム、及び記録媒体 |
Family Cites Families (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
GB9825379D0 (en) * | 1998-11-19 | 1999-01-13 | Electrosonic Ltd | Image processing |
US8180180B2 (en) * | 2008-12-29 | 2012-05-15 | Arcsoft Hangzhou Co., Ltd. | Method for magnifying images and videos |
EP2416557A4 (en) | 2009-03-31 | 2013-01-23 | Sharp Kk | IMAGE ENHANCEMENT DEVICE, IMAGE ENHANCEMENT METHOD, IMAGE ENHANCEMENT PROGRAM, AND SIGNAL PROCESSING DEVICE |
EP2615579A1 (en) * | 2012-01-12 | 2013-07-17 | Thomson Licensing | Method and device for generating a super-resolution version of a low resolution input data structure |
EP3065392B1 (en) * | 2013-10-31 | 2018-08-22 | Fujifilm Corporation | Image processing device, image capture device, parameter generating method, image processing method, and program |
-
2015
- 2015-01-13 WO PCT/JP2015/050663 patent/WO2015115168A1/ja active Application Filing
- 2015-01-13 US US15/036,868 patent/US9679355B2/en not_active Expired - Fee Related
- 2015-01-13 JP JP2015559850A patent/JP6158956B2/ja not_active Expired - Fee Related
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2007074307A (ja) * | 2005-09-06 | 2007-03-22 | Nippon Telegr & Teleph Corp <Ntt> | 時間解像度の拡大方法及び装置及びプログラム及びコンピュータ読み取り可能な記録媒体 |
JP2011188478A (ja) * | 2010-02-09 | 2011-09-22 | Panasonic Corp | 超解像処理装置及び超解像処理方法 |
JP2013021635A (ja) * | 2011-07-14 | 2013-01-31 | Sony Corp | 画像処理装置、画像処理方法、プログラム、及び記録媒体 |
Non-Patent Citations (1)
Title |
---|
HIDENOBU HASHIKAMI: "Gazo Shori", IMAGE LABORATORY, vol. 24, no. 4, 10 April 2013 (2013-04-10), pages 23 - 28 * |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2020241337A1 (ja) * | 2019-05-24 | 2020-12-03 | 株式会社日立製作所 | 画像処理装置 |
Also Published As
Publication number | Publication date |
---|---|
JP6158956B2 (ja) | 2017-07-05 |
US9679355B2 (en) | 2017-06-13 |
JPWO2015115168A1 (ja) | 2017-03-23 |
US20160307293A1 (en) | 2016-10-20 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
JP2015507273A (ja) | 低解像入力データ構造の超解像バージョンを生成する方法及び装置 | |
CN102194216B (zh) | 图像处理设备和图像处理方法 | |
Hung et al. | Fast image interpolation using the bilateral filter | |
JP2013235594A (ja) | 低解像度の入力データ構造の超解像度バージョンを生成する方法及び装置 | |
WO2015180053A1 (zh) | 一种快速超分辨率图像重建方法和装置 | |
JP2011170456A5 (ja) | ||
WO2010140281A1 (ja) | 信号処理装置、信号処理装置の制御方法、制御プログラム、および該制御プログラムを記録したコンピュータ読み取り可能な記録媒体 | |
CN103985085A (zh) | 图像超分辨率放大的方法和装置 | |
CN104700360B (zh) | 基于边缘自适应的图像缩放方法及系统 | |
Mahajan et al. | Adaptive and non-adaptive image interpolation techniques | |
JP4847531B2 (ja) | 画像処理装置、画像処理プログラム、および画像処理方法 | |
WO2016075914A1 (ja) | 画像信号処理装置、画像信号処理方法及び画像信号処理プログラム | |
JP6158956B2 (ja) | 画像処理装置 | |
Xiao et al. | Detail enhancement of image super-resolution based on detail synthesis | |
US9996908B2 (en) | Image processing apparatus, image pickup apparatus, image processing method, and non-transitory computer-readable storage medium for estimating blur | |
Jindal et al. | A novel image zooming technique using wavelet coefficients | |
Zhao et al. | Single depth image super-resolution with multiple residual dictionary learning and refinement | |
WO2012124591A1 (ja) | 信号処理装置、制御プログラム、および集積回路 | |
JP5388780B2 (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 | |
WO2016131370A1 (zh) | 一种实现小波去噪的方法和装置 | |
Khosravi et al. | A new statistical technique for interpolation of landsat images | |
JP2010257179A (ja) | 画像処理装置及び画像処理プログラム | |
KR101332030B1 (ko) | 영상 확대방법 및 이를 수행하는 프로그램을 기록한 컴퓨터로 읽을 수 있는 기록매체 | |
JP5683307B2 (ja) | 間引きフィルタ及びプログラム | |
JP5750349B2 (ja) | 画像復元装置、画像復元方法およびプログラム |
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: 15743942 Country of ref document: EP Kind code of ref document: A1 |
|
ENP | Entry into the national phase |
Ref document number: 2015559850 Country of ref document: JP Kind code of ref document: A |
|
WWE | Wipo information: entry into national phase |
Ref document number: 15036868 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: 15743942 Country of ref document: EP Kind code of ref document: A1 |