CN106342328B - Electronics racemization method for parallel processing based on TIDSP - Google Patents
Electronics racemization method for parallel processing based on TIDSPInfo
- Publication number
- CN106342328B CN106342328B CN200810075529.4A CN200810075529A CN106342328B CN 106342328 B CN106342328 B CN 106342328B CN 200810075529 A CN200810075529 A CN 200810075529A CN 106342328 B CN106342328 B CN 106342328B
- Authority
- CN
- China
- Prior art keywords
- point
- former
- parallel
- coordinate
- interpolation
- 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.)
- Expired - Fee Related
Links
Abstract
The present invention relates to the electronics racemization method for parallel processing based on TI DSP to image rotation algorithm in electronics racemization application. The present invention, on the basis of TI DSP multi-channel parallel processing, carries out algorithm optimization, has overcome the large shortcoming of traditional racemization method computing, real-time implementation electronics racemization. The present invention has improved traditional Rotation Algorithm flow process, is divided into three sections of flow processs that is:, makes this three sections of flow process parallel processings, is equivalent to realize the only 1/3 original time of needs of rotation.
Description
Technical field
The present invention relates to the electricity based on TI DSP to image rotation algorithm in electronics racemization applicationSub-racemization method for parallel processing.
Background technology
In airborne opto-electronic device, electronics racemization technology can provide all the time the not figure of rotation for pilotPicture, and traditional racemization algorithm operation quantity is large, is difficult to realize in embedded system. Traditional revolvesTurn algorithm thinking as follows:
Fig. 1 coordinate origin is chosen in picture centre, is rotated minimum within the scope of 360 degreeAngle 1 is spent. If coordinate is (x, y) before rotation, rotation recoil is designated as (x1, y1), and the anglec of rotation isθ, rotation formula is derived as follows:
Just obtain the coordinate (x, y) of the source images that postrotational point (x1, y1) is corresponding according to 2 formulas,This coordinate of ordinary circumstance is not integer, in order to obtain the pretty good image rotating of effect, just needs profitCarry out interpolation arithmetic with this coordinate, consider the problem of amount of calculation, select bilinear interpolation.
If the former figure coordinate being obtained by rotation is (x, y), the rounded coordinate that this coordinate is corresponding is(x ', y '), bilinear interpolation is specific as follows:
(x ', y) some gray value f (x ', y)=f (x ', y ')+(y-y ') * [f (x ', y '+1)-f (x ', y ')],
(x '+1, y) some gray value
f(x′+1,y)=f(x′+1,y′)+(y-y′)*[f(x′+1,y′+1)-f(x′+1,y′)]
According to (x ', y) and (x '+1, y) 2 gray values, just can obtain (x, y) put gray value:
f(x,y)=f(x′,y)+(x-x′)*[f(x′+1,y)-f(x′,y)]
F (x, y) is exactly the gray scale of pixel after rotation.
Traditional racemization method needs a large amount of mathematical operations in the time completing calculating, and image data amountGreatly, realizing racemization needs for a long time, is difficult to real-time engineering application.
Summary of the invention
The object of the invention is to propose base short and engineering application in real time time of a kind of racemization timeIn the electronics racemization method for parallel processing of TI DSP. Technical solution of the present invention is: concreteTreatment step is as follows:
1) first point in the output image upper left corner is calculated to former figure coordinate corresponding to each point, also simultaneouslyRow reads the gray scale of 4 neighborhoods of the corresponding former figure coordinate of its previous point and walks abreast previous again to itPoint carries out bilinear interpolation, but owing to being just to have started first point to process, walks abreast 4 that readNeighborhood gray scale and interpolation are all invalid datas, therefore do not record 4 gray scales and the interpolation knot that readReally;
2) by from left to right, order from top to bottom, the former figure that second point is calculated to its correspondence sitsMark, the parallel gray scale that reads the corresponding former Fig. 4 neighborhood of its previous point simultaneously, walk abreast to its again beforeA point carries out bilinear interpolation, and now, owing to only having processed two points, interpolation result is still invalid,Not record;
3) since the 3rd point, data are all effective. The former figure that current point is calculated to its correspondence sitsMark, the parallel gray scale that reads the corresponding former Fig. 4 neighborhood of its previous point simultaneously, walk abreast to its again beforeA point carries out bilinear interpolation, and by interpolation result output, as the result of rotation;
4) change next point, forward the 3rd to) continue successive image processing, until entire image isLatter two point turns the 5th);
5) latter two owing to cannot, by parallel processing above, processing separately. Right separatelyPenultimate point carries out bilinear interpolation;
6) last point is read to the gray scale of its corresponding former Fig. 4 neighborhood and complete bilinearity insertValue.
The advantage that the present invention has and effect
The present invention, on the basis of TI DSP multi-channel parallel processing, carries out algorithm optimization, overcomesThe large shortcoming of traditional racemization method computing, real-time implementation electronics racemization. The present invention has improved biographyThe Rotation Algorithm flow process of system is divided into three sections of flow processs that is:, makes this three sections of flow process parallel processings,Be equivalent to realize the only 1/3 original time of needs of rotation.
The present invention is to image line by line according to from left to right, and order rotation from top to bottom, at meterWhen calculating i the corresponding former figure coordinate of point of the rear image of rotation, calculate according to i-1 pointThe parallel gray value that reads its 4 neighborhoods of former figure coordinate, simultaneously parallel i-2 is rotated afterPoint according to its read 4 neighborhood gray values carry out bilinear interpolation, these three flow processs completely withIn time, processes. The inventive method has fully been optimized algorithm, and a complete rotation, interpolation flow process are cutBe 3 flow processs, parallel processing, shortens to original 1/3 by the processing time; Simultaneously at TI DSPOn basis, utilize its MIMD parallel processing capability, well realized the optimization of this algorithm,The processing speed that has greatly improved electronics racemization, is applied it.
Brief description of the drawings
Fig. 1 is the rotation schematic diagram of prior art of the present invention;
Fig. 2 is prior art bilinear interpolation schematic diagram of the present invention;
Fig. 3 is electronics racemization flow chart of the present invention.
Specific implementation method
1) first point in the output image upper left corner is calculated to former figure coordinate 1 corresponding to each point, simultaneouslyWalk abreast and read the gray scale of 4 neighborhoods of the corresponding former figure coordinate of its previous point and walk abreast last again to itIndividual point carries out bilinear interpolation, but owing to being just to have started first point to process, parallel read 4Individual neighborhood gray scale and interpolation are all invalid datas, therefore do not record 4 gray scales and the interpolation that readAs a result 2;
2) by from left to right, order from top to bottom, the former figure that second point is calculated to its correspondence sitsMark 3, the parallel gray scale 4 that reads the corresponding former Fig. 4 neighborhood of its previous point simultaneously, parallel to itPrevious point carries out bilinear interpolation again, and now, owing to only having processed two points, interpolation result stillInvalid, do not record 5;
3) since the 3rd point, data are all effective. The former figure that current point is calculated to its correspondence sitsMark 6, the parallel gray scale 7 that reads the corresponding former Fig. 4 neighborhood of its previous point simultaneously, parallel to itPrevious point carries out bilinear interpolation again, and by interpolation result output, as the result 8 of rotation;
4) change next point, forward the 3rd to) continue successive image to process 9, until entire imageLatter two point turns the 5th);
5) latter two owing to cannot, by parallel processing above, processing separately. Right separatelyPenultimate point carries out bilinear interpolation 10, and last point is read to its corresponding former figure simultaneouslyThe gray scale 11 of 4 neighborhoods;
6) complete the bilinear interpolation 12 of last point.
Embodiment: illustrate the application of racemization algorithm after improving below.
In actual applications, input image resolution 720 × 576, the flow process according to after improving: rightFirst point calculates corresponding former figure coordinate, now inserting of parallel 4 neighborhood gray scales that read and calculatingValue is all invalid data, not record; Second point calculated to corresponding former figure coordinate, before parallel readingThe gray scale of a corresponding former Fig. 4 neighborhood of point, the interpolation result now calculating is invalid, not record.Since the 3rd point, all data are effective, i point calculated to the former figure coordinate of its correspondence,Read the gray scale of 4 neighborhoods of i-1 point simultaneously, and i-2 point carried out to interpolation simultaneously,And result is exported. To the last separately process separately when two somes. Three step executed in parallel,Be equivalent to if the original processing time 1/3, in application, process that an image slices vegetarian refreshments takiesMachine cycle number is to be 1.7 nanoseconds in 24, the DSP cycle. Having calculated like this piece image revolvesThe time turning is 720 × 576 × 24 × 1.7=17 millisecond. The frame frequency of realtime graphic is 40 milliseconds, because ofMethod after this improvement is well positioned to meet practical application.
Claims (1)
1. the electronics racemization method for parallel processing based on TI DSP, is characterized in that, presses from a left sideTurn right, order from top to bottom,
1) first point in the output image upper left corner is calculated, and calculate the former figure of this some correspondenceCoordinate, walks abreast simultaneously and reads the gray scale of 4 neighborhoods of the corresponding former figure coordinate of its previous point and walk abreastTo its more previous point carry out bilinear interpolation, but 4 gray scales that read and not record of interpolation resultStep;
2) second point calculated to the former figure coordinate of its correspondence, simultaneously parallel to read its previous point rightGray scale that should former Fig. 4 neighborhood, parallel to its more previous point carry out bilinear interpolation, now slottingValue result is still invalid, non-registering step;
3) change next point, current point is calculated to the former figure coordinate of its correspondence, walk abreast and read it simultaneouslyThe gray scale of the corresponding former Fig. 4 neighborhood of previous point, parallel to its more previous point carry out bilinearity and insertValue, and by interpolation result output, as the step of the result of rotating;
4) change next point, current point is calculated to the former figure coordinate of its correspondence, walk abreast and read it simultaneouslyThe gray scale of the corresponding former Fig. 4 neighborhood of previous point, parallel to its more previous point carry out bilinearity and insertValue, and by interpolation result output, as the step of the result of rotating; Until entire image is lastTwo points turn the 5th) step;
5) independent step of penultimate point being carried out to bilinear interpolation;
6) last point is read to the gray scale of its corresponding former Fig. 4 neighborhood and complete bilinearity insertThe step of value.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN200810075529.4A CN106342328B (en) | 2008-05-23 | 2008-05-23 | Electronics racemization method for parallel processing based on TIDSP |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN200810075529.4A CN106342328B (en) | 2008-05-23 | 2008-05-23 | Electronics racemization method for parallel processing based on TIDSP |
Publications (1)
Publication Number | Publication Date |
---|---|
CN106342328B true CN106342328B (en) | 2012-07-25 |
Family
ID=58359017
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN200810075529.4A Expired - Fee Related CN106342328B (en) | 2008-05-23 | 2008-05-23 | Electronics racemization method for parallel processing based on TIDSP |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106342328B (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109658337A (en) * | 2018-11-21 | 2019-04-19 | 中国航空工业集团公司洛阳电光设备研究所 | A kind of FPGA implementation method of image real-time electronic racemization |
CN111667413A (en) * | 2020-05-27 | 2020-09-15 | 江苏和正特种装备有限公司 | Image despinning method and system based on multi-source sensing data fusion processing |
CN113962842A (en) * | 2021-10-20 | 2022-01-21 | 北京航空航天大学 | Dynamic stepless despinning system and method based on large-scale integrated circuit high-level synthesis |
-
2008
- 2008-05-23 CN CN200810075529.4A patent/CN106342328B/en not_active Expired - Fee Related
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109658337A (en) * | 2018-11-21 | 2019-04-19 | 中国航空工业集团公司洛阳电光设备研究所 | A kind of FPGA implementation method of image real-time electronic racemization |
CN109658337B (en) * | 2018-11-21 | 2023-03-24 | 中国航空工业集团公司洛阳电光设备研究所 | FPGA implementation method for real-time electronic despinning of images |
CN111667413A (en) * | 2020-05-27 | 2020-09-15 | 江苏和正特种装备有限公司 | Image despinning method and system based on multi-source sensing data fusion processing |
CN113962842A (en) * | 2021-10-20 | 2022-01-21 | 北京航空航天大学 | Dynamic stepless despinning system and method based on large-scale integrated circuit high-level synthesis |
CN113962842B (en) * | 2021-10-20 | 2022-12-09 | 北京航空航天大学 | Dynamic non-polar despinning system and method based on high-level synthesis of large-scale integrated circuit |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Banz et al. | Real-time stereo vision system using semi-global matching disparity estimation: Architecture and FPGA-implementation | |
CN101271196B (en) | Lens shade correction index confirming method, lens shade emendation method and device | |
CN110223232A (en) | A kind of video image amplifying method based on bilinear interpolation algorithm | |
CN106342328B (en) | Electronics racemization method for parallel processing based on TIDSP | |
CN107239216A (en) | Drawing modification method and apparatus based on touch-screen | |
KR20110062083A (en) | Video restoration apparatus and its method | |
CN103679776A (en) | Image processing apparatus | |
CN111523652A (en) | Processor, data processing method thereof and camera device | |
CN111861868B (en) | Image processing method and device for beautifying human images in video | |
Chalimbaud et al. | Embedded active vision system based on an FPGA architecture | |
JP2947360B2 (en) | Method and apparatus for measuring image motion | |
CN116070687A (en) | Neural network light field representation method based on global ray space affine transformation | |
CN114463178A (en) | Image processing chip, method and equipment | |
CN105550986B (en) | A kind of image mapping methods and device based on more camera lens multisensors | |
CN102663783B (en) | Image realization apparatus and image realization method thereof | |
CN105513011B (en) | Image-scaling method | |
CN108510533B (en) | Fourier mellin registration and Laplace fusion image acceleration system based on FPGA | |
Mori et al. | An unified approach for convolution-based image filtering on reconfigurable systems | |
Jia et al. | A miniature stereo vision machine for real-time dense depth mapping | |
CN114581548A (en) | Scene model surface texture superposition method and device based on collision detection | |
CN112150387B (en) | Method and device for enhancing stereoscopic impression of five sense organs on human images in photo | |
CN112580492A (en) | Vehicle detection method and device | |
CN112184833B (en) | Hardware implementation system and method of binocular stereo matching algorithm | |
Benabid et al. | FPGA-based real-time embedded vision system for autonomous mobile robots | |
CN109785241A (en) | A kind of distortion correction method and device, a kind of depth perception calculation method and device |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
GR03 | Grant of secret patent right | ||
DC01 | Secret patent status has been lifted | ||
DCSP | Declassification of secret patent | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20120725 Termination date: 20180523 |
|
CF01 | Termination of patent right due to non-payment of annual fee |