CN110349100B - Proportional pixel extraction method along fuzzy path - Google Patents
Proportional pixel extraction method along fuzzy path Download PDFInfo
- Publication number
- CN110349100B CN110349100B CN201910558377.1A CN201910558377A CN110349100B CN 110349100 B CN110349100 B CN 110349100B CN 201910558377 A CN201910558377 A CN 201910558377A CN 110349100 B CN110349100 B CN 110349100B
- Authority
- CN
- China
- Prior art keywords
- pixel
- arc
- sequence
- image
- fuzzy
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000605 extraction Methods 0.000 title claims abstract description 15
- 238000000034 method Methods 0.000 claims abstract description 31
- 230000033001 locomotion Effects 0.000 claims abstract description 25
- 230000008569 process Effects 0.000 claims abstract description 11
- 230000003247 decreasing effect Effects 0.000 claims description 2
- 238000009795 derivation Methods 0.000 claims description 2
- 238000012545 processing Methods 0.000 description 5
- 238000004422 calculation algorithm Methods 0.000 description 4
- 238000004364 calculation method Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 4
- 230000015556 catabolic process Effects 0.000 description 2
- 238000006731 degradation reaction Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000003384 imaging method Methods 0.000 description 2
- 238000011084 recovery Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 238000007734 materials engineering Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 230000006641 stabilisation Effects 0.000 description 1
- 238000011105 stabilization Methods 0.000 description 1
- 230000002123 temporal effect Effects 0.000 description 1
- 238000000844 transformation Methods 0.000 description 1
- 238000013519 translation Methods 0.000 description 1
Images
Classifications
-
- 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
-
- 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
- G06T2207/10—Image acquisition modality
- G06T2207/10016—Video; Image sequence
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Image Processing (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a proportional pixel extraction method along a fuzzy path, which comprises the steps of firstly inputting a fuzzy image, utilizing an eighth circle drawing method on the fuzzy image, taking an eighth fuzzy pixel sequence on an eighth circular arc with x being more than or equal to 0 and less than or equal to y, and obtaining a radius ofrOn the eighth circle of (C) a series ofMA sequence of dots, a series ofMThe sequence of points is transformed by mirror rotation coordinates to obtain a radius ofrOver the entire circumference ofMThe sequence of points is formed to obtain fuzzy pixel sequences of all radii, the fuzzy pixel sequences are subjected to image restoration by a one-dimensional translational motion fuzzy restoration method, and the radii are changedrAnd performing one-dimensional image restoration on all the pixel sequences with different radiuses to obtain restored pixel sequences with all the radiuses, performing the inverse process of the flow on the restored pixel sequences with all the radiuses to obtain a backfilled coordinate corresponding relation, and backfilling the restored pixel sequences according to the corresponding relation to obtain a restored clear image.
Description
Technical Field
The invention belongs to the image processing technology, and particularly relates to a proportional pixel extraction method along a fuzzy path.
Background
In recent years, under another generation of digital wave raised by artificial intelligence, a new vitality is revived in the field of image processing, and in the processes of photoelectric conversion, transmission, recording and displaying of an image, because a scene and a camera may generate a series of spatial motions when an actual imaging system shoots, energy is abnormally accumulated on an imaging plane in the process of one exposure, and the obtained image generates degradation blur, which is generally called motion blur. The process of processing the blurred image to recover the original real image as much as possible is called image restoration, which is also called image recovery.
Restoration of a rotational motion blurred image is a special case in the field of image restoration, but is also most common, and can be decomposed into translational motion and rotational motion no matter how complex the motion is. The translational motion has space invariance, the degradation mechanism is simpler, and the fuzzy kernels in each pixel neighborhood are basically consistent, so the estimation of the fuzzy kernels is relatively easy. However, for the rotational blur, the direction of the blur kernel in the neighborhood of each pixel point is the tangential direction of the current pixel on the circle where the current pixel is located, and on the other hand, the size of the blur kernel also varies with the size of the circle where the current pixel is located, and has no spatial variability. Therefore, the estimation of the blur kernel for rotational motion blur is quite complex.
From daily life photographing to laboratory photographing of high-speed moving targets, restoration of motion blurred images is widely applied to many fields such as optical anti-shake, machine vision, video monitoring and the like. Because the image blurring phenomenon generated by the motion can cause difficulty to the later work, the method is also the basis of the follow-up work of pattern recognition, image registration, video image sequence stabilization and the like, and has very important research value. How to find a rotational motion blur restoration algorithm with ideal restoration effect and good real-time performance is an important work in the image processing process and is also a significant subject.
For the spatial variability characteristic of rotational motion blur, scholars at home and abroad have developed intensive research on how to change rotational motion blur into spatial invariant blur:
a solution based on geometric transformation is proposed in Space-Variant Image retrieval by Coordinate Transformations (J.Opt.Soc.Amer.,64(2), pp.138-144, February, 1974) of A.A.Sawchuk and M.J.Peyrobian, which directly transforms rectangular coordinates of an Image to polar coordinates, and then interpolates gray values of the rest coordinates by a gray level interpolation method, so that the rotation blur is converted into the ordinary translation motion blur for processing.
Zhou Jianyang, Yang Yintao, Wu Yiliang in Fast recovery algorithm for temporal movement of blurred images Y, an International Conference on Materials Engineering for Advanced Technologies, ICMEAT 2011, proposes a method for extracting pixel sequences along a fuzzy path based on a Bresenham circle drawing method, which improves the real-time performance of the algorithm.
Disclosure of Invention
The invention aims to provide a proportional pixel extraction method along a fuzzy path, which has the characteristics of small calculation amount and no loss of image information, can effectively extract pixel points along a rotational motion fuzzy path and restore a rotational motion fuzzy image.
The technical solution for realizing the purpose of the invention is as follows: a proportional pixel extraction method along a fuzzy path comprises the following steps:
step 1, inputting a blurred image, and turning to step 2;
step 2, on the blurred image, taking the center of the blurred image as the circle center, taking the radius as r, taking the X axis as a horizontal axis parallel to the paper surface, taking the Y axis as a vertical axis parallel to the paper surface, taking an eighth blurred pixel sequence on an eighth arc of which X is more than or equal to 0 and less than or equal to Y by utilizing an eighth circle drawing method, and selecting any point P on the eighth arc i (x i ，y i ) Then, four reference points H1 (x) are selected i +1，y i )、H2(x i +1，y i +1)、L1(x i +1，y i -1)、L2(x i +1，y i -2), the eighth arc must pass between two adjacent reference points of the four reference points, the pixel value of the intersection point M of the line connecting the two adjacent reference points and the eighth arc is calculated, and the reference point closest to the eighth arc is updated to P i Selecting a corresponding reference point, repeating the steps to obtain a sequence consisting of a series of M points on an eighth circle with the radius r, turning to the step 3,
step 3, converting a series of sequences consisting of M points on an eighth arc with the radius r by using mirror image rotation coordinates to obtain a sequence consisting of M points on the whole circumference with the radius r, calling the sequence as a fuzzy pixel sequence of an image to obtain fuzzy pixel sequences of all radii, and turning to step 4;
step 4, performing image restoration on the blurred pixel sequence by a one-dimensional translational motion blur restoration method to obtain a restored pixel sequence, and turning to step 5;
step 5, changing the radius r, returning to the step 2, performing one-dimensional image restoration on all the pixel sequences with different radii to obtain restored pixel sequences with all the radii, and turning to the step 6;
and 6, backfilling the restored pixel sequences according to the backfilled coordinate corresponding relation obtained by the restored pixel sequences with all radii according to the inverse process of the proportional pixel extraction method along the fuzzy path in the steps 2 and 3, and backfilling the restored pixel sequences according to the corresponding relation to obtain the restored clear image.
Compared with the prior art, the invention has the remarkable advantages that:
(1) the image deblurring method along the fuzzy path is utilized to deblur the image on the blurred image generated when the camera and the scenery do relative rotation motion, the defect that high-frequency information is lost in the traditional rounding method is overcome, the image information is fully utilized, and secondary image gray level interpolation is not needed.
(2) The signal-to-noise ratio of the restored image is remarkably improved under the condition of small calculation amount, and the peak signal-to-noise ratio (PSNR) of the 60-degree rotation blurred image is improved to 25.02dB from 16.79dB by the algorithm.
Drawings
FIG. 1 is a flow chart of a proportional pixel extraction method along a fuzzy path according to the present invention.
Fig. 2 is a schematic diagram of the rotational symmetry of a circle.
FIG. 3 is a schematic diagram of determining an i +1 point by the proportional pixel extraction method along the blur path according to the present invention.
Fig. 4 shows a motion blurred image rotated by 60 degrees within the exposure time according to the present invention.
FIG. 5 is a sequence diagram of blurred pixels extracted from a rotational motion blurred image using a proportional pixel extraction method along a blur path according to the present invention.
FIG. 6 is a diagram of an image reconstructed from a rotational motion blur image using a proportional pixel extraction along a blur path according to the present invention.
Detailed Description
The present invention is described in further detail below with reference to the attached drawing figures.
With reference to fig. 1, the method for extracting a proportional pixel along a fuzzy path according to the present invention includes the following steps:
step 1, inputting a blurred image as shown in fig. 4, and turning to step 2.
Step 2, combining with the graph 2, on the fuzzy image, taking the center of the fuzzy image as the center of a circle, the radius of the fuzzy image is r, the X axis is a horizontal axis parallel to the paper surface, the Y axis is a vertical axis parallel to the paper surface, utilizing an eighth circle drawing method to take an eighth fuzzy pixel sequence on an eighth circular arc with X being more than or equal to 0 and less than or equal to Y, combining with the graph 3, selecting any point P on the eighth circular arc i (x i ，y i ) Then, four reference points H1 (x) are selected i +1，y i )、H2(x i +1，y i +1)、L1(x i +1，y i -1)、L2(x i +1，y i -2), the eighth arc must pass between two adjacent reference points of the four reference points, the pixel value of the intersection point M of the line connecting the two adjacent reference points and the eighth arc is calculated, and the reference point closest to the eighth arc is updated to P i And (3) selecting a corresponding reference point, repeating the steps to obtain a sequence consisting of a series of M points on an eighth circle with the radius r, and turning to the step 3, wherein the sequence specifically comprises the following steps:
step 2-1, utilizing an eighth circle drawing method to obtain an eighth fuzzy pixel sequence on an eighth circular arc with x being more than or equal to 0 and less than or equal to y, and enabling x to be i ＝0，y i R, and the intersection point of the Y axis and the arc is taken as P i (0, r), the four reference points are respectively H1(1, r), H2(1, r +1), L1(1, r-1) and L2(1, r-2), and the process goes to step 2-2.
Step 2-2, selecting two reference points closest to one eighth of arc and calculating the proportion q of the distance between the M point and the two reference points i It can be calculated by the following formula:
where dH is the distance from point M to point H1; turning to the step 2-3;
step 2-3, combine with FIG. 3, when P i (x i ，y i ) X is more than or equal to 0On the one eighth arc less than or equal to y, because: the value of y is monotonically decreased, and the analytical expression of the circle at this time is
Derivation of this can yield:
the slope e-1, 0 of the available arc]That is, two reference points closest to the arc are necessarily H1 or L1, and if one-eighth of the arc passes through the reference points, q is determined as follows i To which case it belongs:
1) if q is i <0 (pixel), then the eighth arc passes through the two reference points H1 and H2 and is nearest to H1, and step 2-4 is carried out;
2) if 0 (pixel) is less than or equal to q i 0.5 (pixel) or less, then the one-eighth arc passes through two reference points H1 and L1 and is nearest to H1, and the step 2-4 is carried out;
3) if 0.5 (Pixel)<q i 1 (pixel), then the one-eighth arc crosses both H1 and L1 reference points and is closest to L1, proceed to step 2-4;
4) if 1 (Pixel)<q i Then the eighth arc crosses both reference points L1 and L2 and is closest to L1, and proceeds to step 2-4.
Step 2-4, two reference points which are penetrated by one eighth of circular arc are proportioned to q i Calculating the pixel value of M point, and updating the reference point closest to the eighth arc to P i And (3) selecting a corresponding reference point, repeating the steps 2-2 to 2-3 to obtain a sequence consisting of a series of M points on an eighth circle with the radius r, and turning to the step 3.
In the four cases of step 2-3, if the arc is close to point H1, the ratio q of the next M points i+1 The simplified formula may be utilized:
calculating; if the arc is close to point L1, the ratio q of the next M points i+1 The simplified formula may be utilized:
and (4) calculation, the calculation amount can be reduced.
And 3, converting the sequence formed by the series of M points on the one-eighth arc with the radius of r obtained in the step 2-4 by using mirror image rotation coordinates to obtain a sequence formed by the M points on the whole circumference with the radius of r, referring the sequence to be a blurred pixel sequence of the image, obtaining the blurred pixel sequences with all radii as shown in the figure 5, and turning to the step 4.
And 4, performing image restoration (such as wiener filtering) on the blurred pixel sequence by a one-dimensional translational motion blur restoration method to obtain a restored pixel sequence, and turning to the step 5.
And 5, changing the radius r, returning to the step 2, performing one-dimensional image restoration on all the pixel sequences with different radii to obtain restored pixel sequences with all the radii, and turning to the step 6.
And 6, backfilling the restored pixel sequences according to the backfilled coordinate corresponding relation obtained by the restored pixel sequences with all radii according to the inverse process of the proportional pixel extraction method along the fuzzy path in the steps 2 and 3, and obtaining a restored clear image as shown in the figure 6.
Claims (3)
1. A proportional pixel extraction method along a fuzzy path is characterized by comprising the following steps:
step 1, inputting a blurred image, and turning to step 2;
step 2, on the blurred image, taking the center of the blurred image as the circle center, taking the radius as r, taking the X axis as a horizontal axis parallel to the paper surface, taking the Y axis as a vertical axis parallel to the paper surface, taking an eighth blurred pixel sequence on an eighth arc of which X is more than or equal to 0 and less than or equal to Y by utilizing an eighth circle drawing method, and selecting any point P on the eighth arc i (x i ，y i ) Then, four reference points H1 (x) are selected i +1，y i )、H2(x i +1，y i +1)、L1(x i +1，y i -1)、L2(x i +1，y i -2) an eighth arc must pass between two adjacent ones of said four reference pointsCalculating the pixel value of the intersection point M of the connecting line between two adjacent reference points and the eighth circular arc, and updating the reference point closest to the eighth circular arc to be P i Selecting a corresponding reference point, and repeating the steps to obtain a sequence consisting of a series of M points on an eighth circle with the radius r, wherein the sequence comprises the following specific steps:
step 2-1, utilizing an eighth circle drawing method to obtain an eighth fuzzy pixel sequence on an eighth circular arc with x being more than or equal to 0 and less than or equal to y, and enabling x to be i ＝0，y i R, and the intersection point of the Y axis and the arc is taken as P i (0, r), the four reference points are respectively H1(1, r), H2(1, r +1), L1(1, r-1) and L2(1, r-2), and the process is carried out to step 2-2;
step 2-2, selecting two reference points closest to one eighth of arc and calculating the proportion q of the distance between the M point and the two reference points i Calculated from the following equation:
where dH is the distance from point M to point H1; turning to the step 2-3;
step 2-3, when P i (x i ，y i ) On the one eighth arc where x is more than or equal to 0 and less than or equal to y, the following factors are used: the value of y is monotonically decreased, and the analytical expression of the circle at this time is
The derivation of which is:
whereby the slope e-1, 0 of an eighth-degree arc]That is, two reference points closest to the eighth arc must be H1 or L1, and thus the eighth arc passes through the reference points in the following four cases, and q is determined i To which case it belongs:
1) if q is i <Pixel 0, then one eighth of the circular arc passes through two reference points H1 and H2 and is nearest to H1, and the process goes to step 2-4;
2) if 0 pixel is less than or equal to q i If the pixel number is less than or equal to 0.5, an eighth circular arc passes through two reference points of H1 and L1 and is nearest to H1, and the step 2-4 is carried out;
3) if 0.5 pixel<q i If the pixel number is less than or equal to 1 pixel, an eighth circular arc passes through two reference points of H1 and L1 and is nearest to L1, and the step 2-4 is carried out;
4) if 1 pixel<q i If the eighth circular arc passes through two reference points of L1 and L2 and is closest to L1, the step 2-4 is carried out;
step 2-4, two reference points which are penetrated by one eighth of circular arc are proportioned to q i Calculating the pixel value of M point, and updating the reference point closest to the eighth arc to P i Selecting a corresponding reference point, repeating the steps 2-2 to 2-3, and obtaining a sequence consisting of a series of M points on an eighth circle with the radius r;
the procedure is shifted to the step 3,
step 3, converting a series of sequences consisting of M points on an eighth arc with the radius r by using mirror image rotation coordinates to obtain a sequence consisting of M points on the whole circumference with the radius r, calling the sequence as a fuzzy pixel sequence of an image to obtain fuzzy pixel sequences of all radii, and turning to step 4;
step 4, performing image restoration on the blurred pixel sequence by a one-dimensional translational motion blur restoration method to obtain a restored pixel sequence, and turning to step 5;
step 5, changing the radius r, returning to the step 2, performing one-dimensional image restoration on all the pixel sequences with different radii to obtain restored pixel sequences with all the radii, and turning to the step 6;
and 6, obtaining a backfilled coordinate corresponding relation of the restored pixel sequences with all radii according to the inverse process of the step 2 and the step 3, and backfilling the restored pixel sequences according to the corresponding relation to obtain the restored clear image.
2. The method of proportional pixel extraction along a blur path of claim 1, wherein: in the four cases of step 2-3, if the arc is close to point H1, the ratio q of the next M points i+1 Using a simplified formula:
and (4) calculating.
3. The method of scaled pixel extraction along a blur path of claim 1, wherein: in the four cases of step 2-3, if the arc is close to point L1, the ratio q of the next M points i+1 Using a reduction formula:
and (4) calculating.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910558377.1A CN110349100B (en) | 2019-06-26 | 2019-06-26 | Proportional pixel extraction method along fuzzy path |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910558377.1A CN110349100B (en) | 2019-06-26 | 2019-06-26 | Proportional pixel extraction method along fuzzy path |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110349100A CN110349100A (en) | 2019-10-18 |
| CN110349100B true CN110349100B (en) | 2022-08-12 |
Family
ID=68183109
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910558377.1A Active CN110349100B (en) | 2019-06-26 | 2019-06-26 | Proportional pixel extraction method along fuzzy path |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110349100B (en) |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105976332B (en) * | 2016-05-03 | 2019-03-01 | 北京大学深圳研究生院 | Image deblurring method based on bright fringes information in image |
| CN109272456A (en) * | 2018-07-25 | 2019-01-25 | 大连理工大学 | The blurred picture high-precision restoring method of view-based access control model prior information |
-
2019
- 2019-06-26 CN CN201910558377.1A patent/CN110349100B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN110349100A (en) | 2019-10-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109102462B (en) | Video super-resolution reconstruction method based on deep learning | |
| Zheng et al. | Crossnet: An end-to-end reference-based super resolution network using cross-scale warping | |
| Dudhane et al. | Burst image restoration and enhancement | |
| Kappeler et al. | Video super-resolution with convolutional neural networks | |
| CN111028150B (en) | Rapid space-time residual attention video super-resolution reconstruction method | |
| WO2020015167A1 (en) | Image super-resolution and non-uniform blur removal method based on fusion network | |
| CN110599400B (en) | EPI-based light field image super-resolution method | |
| CN111915484A (en) | Reference image guiding super-resolution method based on dense matching and self-adaptive fusion | |
| Song et al. | An adaptive l 1–l 2 hybrid error model to super-resolution | |
| CN109767389B (en) | Self-adaptive weighted double-norm remote sensing image blind super-resolution reconstruction method based on local and non-local combined prior | |
| Tan et al. | Crossnet++: Cross-scale large-parallax warping for reference-based super-resolution | |
| Dudhane et al. | Burstormer: Burst image restoration and enhancement transformer | |
| Yoshikawa et al. | Super resolution image reconstruction using total variation regularization and learning-based method | |
| Luo et al. | Image demosaicing based on generative adversarial network | |
| CN110852947B (en) | Infrared image super-resolution method based on edge sharpening | |
| Mehta et al. | Gated multi-resolution transfer network for burst restoration and enhancement | |
| Tang et al. | Single-image super-resolution based on semi-supervised learning | |
| Tseng et al. | Depth image super-resolution via multi-frame registration and deep learning | |
| Chen et al. | Guided dual networks for single image super-resolution | |
| CN110349100B (en) | Proportional pixel extraction method along fuzzy path | |
| Oh et al. | Fpanet: Frequency-based video demoireing using frame-level post alignment | |
| Li et al. | Spatio-temporal fusion network for video super-resolution | |
| CN112348745B (en) | Video super-resolution reconstruction method based on residual convolutional network | |
| Li et al. | Single image deblurring using bi-attention network | |
| Noor et al. | Multi-frame super resolution with deep residual learning on flow registered non-integer pixel images |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |