CN101853524A - Method for generating corn ear panoramic image by using image sequence - Google Patents
Method for generating corn ear panoramic image by using image sequence Download PDFInfo
- Publication number
- CN101853524A CN101853524A CN201010176315A CN201010176315A CN101853524A CN 101853524 A CN101853524 A CN 101853524A CN 201010176315 A CN201010176315 A CN 201010176315A CN 201010176315 A CN201010176315 A CN 201010176315A CN 101853524 A CN101853524 A CN 101853524A
- Authority
- CN
- China
- Prior art keywords
- image
- image sequence
- pixel
- corn ear
- field picture
- 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.)
- Pending
Links
Images
Landscapes
- Image Processing (AREA)
Abstract
The invention provides a method for generating a corn ear panoramic image by using an image sequence, comprising the following steps: using a position-fixed digital camera to shoot the image sequence of a corn ear at a certain angel interval; extracting an image characteristic point, calculating motion parameters of adjacent two frame images in the image sequence according to the matched characteristic point, and registering the motion parameters to one coordinate system to compose a new image; searching a suture line in the new image, cutting the adjacent two frame images according to the position of the suture line and eliminating the pixel exposure difference at the two sides of the suture line by Gaussian template filtration; stretching the image at the horizontal and vertical directions to counteract perspective projection deformation; and sequentially performing the above operations on the image sequence to synthesize the corn ear panoramic image. The method of the invention can display the seamless splicing overall perspective of the image sequence of the corn ear on one new image, thus greatly reducing the redundancy of picture information, reducing information storage space and enabling the morphologic observation to be more visual.
Description
Technical field
The present invention relates to the computer graphics techniques field, particularly a kind of method of using the image sequence splicing to generate the corn ear panorama sketch.
Background technology
The corn ear outer shape is an important indicator of carrying out species test and the reference of breeding of new variety institute, and one of record profile characteristics the best way is photographic images.Because corn ear is cylindrical, piece image often can not write down full detail, has only around fruit ear rotation shooting multiple image to cover its whole surface.In order to obtain desirable observing effect, a fruit ear needs 5-10 width of cloth image recording usually.Certainly exist lap between several fruit ear images, these laps are difficult to differentiate by naked eyes, and this will cause the inconvenience of information redundancy and observation.
Can adopt full shot (panoramic photography) technology to generate panorama sketch, this technology mainly obtains the wide cut image of subject by specific hardware and software, multi-angle multi-view image sequence is synthesized on piece image, it can be realized by the panorama camera with rotatable camera lens, but this class hardware device price is higher and need certain professional operative knowledge, and the scope of application is restricted.The another kind of method that generates panorama sketch is the synthetic panorama sketch of common imaging device being taken by image mosaic software of image sequence, if prior calibrating camera confidential reference items, around the object photographic images time under the also known situation of camera motion information, image sequence can be projected on the same cylinder in turn, be spliced into complete cylindricality panorama sketch behind the removal lap.But under some application scenario, determine that the camera motion difficulty is bigger, the practical application complexity is loaded down with trivial details.
Generate the technology that the corn ear panorama sketch relates to aspects such as image motion model simplification, figure cutting, elimination difference in exposure, distortion stretching, the solution that does not up to the present still have a cover to be fit to.
Summary of the invention
(1) technical matters that will solve
The technical problem to be solved in the present invention is to overcome in the prior art that the information redundance that exists in the method for using several electronic photos record corn ear external morphologies is big, storage cost is high, be not easy to defective such as observation, and is easy, generate the good corn ear panorama sketch of observing effect apace.
(2) technical scheme
For solving the problems of the technologies described above, technical scheme of the present invention provides a kind of method according to image sequence generation corn ear panorama sketch, may further comprise the steps:
S1: obtain the image sequence that a plurality of images of corn ear are formed by the rotation corn and in the mode that at interval predetermined angle is taken continuously, wherein per two frame adjacent images have the overlapping region;
S2: extract the target image sequence of the fruit ear in the described image sequence by the background subtraction point-score, and it is carried out gray processing handle, obtain the target grayscale image sequence of fruit ear;
S3: by perspective projection transformation the 1st frame in the target gray level image sequence of described fruit ear and the 2nd two field picture are registered to the same coordinate system and form a frame new images I;
S4: search for the best suture line of described new images I by dynamic programming algorithm, cut described the 1st frame and the 2nd two field picture, the right side section of the left side section of described the 1st two field picture and the 2nd two field picture is combined as part panoramic picture I ' along described best suture line;
S5: the image that is positioned at described best suture line left side among the described part panoramic picture I ' is carried out level and stretched vertically,, finish once splicing to eliminate the distortion of described part panoramic picture I ' in the projective transformation process;
S6: to the 3rd two field picture execution in step S3-S5 in the target gray level image sequence of part panoramic picture I ' after the stretch processing and described fruit ear to splice next time, so be repeated to the last frame image in the target gray level image sequence of described fruit ear, obtain the corn ear panorama sketch.
Further, described step S3 comprises following substep:
S301: extract the unique point in the overlapping region of described the 1st frame and the 2nd two field picture by harris, SUSAN or SIFT method;
S302: according to proximity search algorithm described unique point is mated, obtain the many groups of corresponding match point X and the X ' of two two field pictures;
S303: calculate the described homography matrix H that organizes between corresponding match point X and the X ' by least square method more, wherein, X '=HX;
S304: according to described homography matrix H the 1st two field picture is carried out perspective projection transformation, it is registered to the 2nd two field picture place coordinate system, form new images I;
Wherein, the pixel of described new images I is the average of the 1st two field picture and the 2nd two field picture respective pixel after the projective transformation.
Further, also comprise between described step S302 and the S303: remove the described mistake match point of organizing in the corresponding match point by the random sampling consistency algorithm more.
Further, among the described step S4, seek best sutural process by dynamic programming and comprise:
S401: the first row pixel by described new images I begins, and each is listed as first pixel and is designated as a sutural end points;
S402: begin to search for the next line pixel 3 pixels of adjacency from each sutural end points, the pixel of getting the gray-scale value minimum is as described sutural bearing of trend, and search downwards is until the bottom of described new images I line by line;
S403: repeat above-mentioned steps and obtain many suture lines, select a wherein best suture line of conduct of grey scale pixel value sum minimum.
Further, among the described step S4, cut described the 1st frame and the 2nd two field picture along described best suture line before, also eliminate the difference in exposure of described best suture line both sides image by laplacian pyramid blending algorithm or weights blending algorithm.
Wherein, among the described step S5, the pixel-shift amount d of described level and stretched vertically follows following formula:
Wherein d ' is the maximum offset of this row pixel on the draw direction, x
eBe this row pixel boundary coordinate after the projective transformation, x
sBe the coordinate of best suture line at this row, x is for desiring the pixel coordinate of stretch position.
Further, among the step S5, after eliminating the distortion of described part panoramic picture I ' in the projective transformation process, also pixel moves the cavity that is produced in the drawing process to fill up to carrying out bilinear interpolation through the pixel of the described part panoramic picture I ' of level and stretched vertically.
Wherein, among the described step S1, the area of overlapping region is greater than 50% of single-frame images area between described adjacent two two field pictures.
(3) beneficial effect
Several electronic photos that the present invention will write down the corn ear external morphology are spliced into a seamless panorama sketch, thereby have reduced the redundance of photographic intelligence greatly, reduce the information stores space, make morphologic observation more directly perceived; The Machine Vision Detection of also investigating (as: tassel row number, grain number per spike, spike length, fringe diameter, bald sharp rate) for the corn ear formalness provides technical foundation simultaneously.
Description of drawings
Fig. 1 is the process flow diagram that generates the method for corn ear panorama sketch according to use image sequence of the present invention;
Fig. 2 is a process flow diagram according to an embodiment of the invention;
Fig. 3 is according to sutural synoptic diagram in the method for use image sequence generation corn ear panorama sketch of the present invention;
Fig. 4 is the synoptic diagram according to image mosaic in the method for use image sequence generation corn ear panorama sketch of the present invention;
Fig. 5 is the corn ear panorama sketch that generates the method generation of corn ear panorama sketch according to use image sequence of the present invention; Wherein, (a) be a frame fruit ear image in the image sequence, the splicing of two two field pictures is merged before and after (b) showing, and (c) the fruit ear panorama sketch for not carrying out stretch processing (d) is the fruit ear panorama sketch after the stretch processing.
Embodiment
The method according to image sequence generation corn ear panorama sketch that the present invention proposes is described as follows in conjunction with the accompanying drawings and embodiments.
Fig. 1 is the process flow diagram of the method according to this invention, and as shown in Figure 1, the present invention mainly may further comprise the steps:
S1: take the fruit ear rotation diagram; By obtain the image sequence of corn ear along the corn outer surface interval continuous mode of taking of predetermined angular, wherein per two frame adjacent images have the overlapping region;
S2: image pre-service; Extract fruit ear target image sequence in the described image sequence by the background subtraction point-score, and it is carried out gray processing handle, obtain fruit ear target gray level image sequence;
S3: match point detects; Extract the unique point in the overlapping region of the 1st frame and the 2nd two field picture in the described image sequence; According to proximity search algorithm described unique point is mated, obtain many groups of corresponding match points of two two field pictures;
S4: calculate homography matrix; Calculate the described homography matrix of organizing between the corresponding match point by least square method more;
S5: image cutting, stitching; According to the homography matrix of trying to achieve front and back two two field pictures are registered to the same coordinate system and form a frame new images; Search the best suture line in the new images, along suture line cutting front and back two two field pictures, combination section component part panorama sketch;
S6: anti-twist distortion; The part panorama sketch is carried out level and stretched vertically to eliminate the distortion in the projective transformation process;
S7: part panorama sketch after the antagonism deformation process and next frame image repeat above-mentioned steps S3-6 and splice, and until handling the last frame image, generate panorama sketch.
To be further elaborated to method of the present invention by a specific embodiment below.As shown in Figure 2, present embodiment may further comprise the steps:
S101: obtain image; Particularly, digital camera is fixed on the tripod, corn ear is positioned over rotary type measures on the platform, control mensuration platform only rotates to an angle (10-20 degree) fruit ear at every turn and takes a two field picture, obtains the corn ear image sequence thus.
In the specific implementation process, the size of the anglec of rotation can influence subsequent operation, specifically, the fruit ear anglec of rotation is crossed the quantity that conference reduces overlapping region characteristic matching point between two two field pictures of front and back, the robustness that very few match point may cause finding the solution homography matrix descends, under extreme case even may obtain error result; When the anglec of rotation was too small, image sequence quantity increased, and this may add the master control program calculated amount, increased the execution time thus; The more important thing is that when stitching image quantity increased, cumulative errors also increased thereupon, this will reduce the images match precision; In addition, it is big that the image perspective distortion also will add up along with increasing of image sequence to become, and this can increase the difficulty of anti-twist distortion in the aftertreatment, reduces observing effect thus.Through repeatedly experimental verification, fruit ear rotates a circle, and to take 12-18 width of cloth image general effect more satisfactory.In the image acquisition procedures, the illumination that should control environment is relatively stable, and front and back two two field picture laps surpass image area 50% in the image sequence.
S102: image pre-service; Particularly, by the foreground picture and the Background of background subtraction point-score difference corn ear, that is, it is poor to do with described fruit ear image sequence and the Background respective pixel that does not comprise fruit ear, extracts the fruit ear foreground image, obtains the fruit ear target image; Target image is carried out morphological erosion, expansion process removal isolated noise point; Therefore, in the specific implementation process, for ease of pre-service, be further noted that the selection of illumination conversion and background in the image acquisition procedures, for reducing cost to increase the applicability of system, image in the embodiment of the invention all obtains under the lamp environment, the monochromatic curtain effect that shooting background is selected and the fruit ear color distortion is bigger is better, the entire image acquisition process should be finished in the short to the greatest extent time, to get rid of the influence that ambient lighting changes, make later stage stitching image exposure effect homogeneous, help the background subtraction point-score simultaneously and generate the fruit ear target image.
S103: images match; Particularly, the unique point in the overlapping region of two two field pictures before and after the extraction; According to the most contiguous criteria match unique point; The consistance of check match point is removed exterior point; Use the homography matrix between the least square method calculating match point; According to trying to achieve homography matrix the former frame image is made perspective projection transformation, front and back two two field pictures are registered in the same coordinate system, get former frame image and the synthetic width of cloth new images of the average of back one two field picture respective pixel after the conversion.
In the specific implementation process, if same point in the corresponding real world of certain pixel value in two two field pictures of front and back, claim that then these two points are corresponding match point, behind the corresponding match point that obtains some, both can resolve the image motion relation, the splicing of two width of cloth adjacent images is carried out according to this image relative motion relation.The concatenation of front and back two two field pictures can be carried out according to following image motion model: and a pair of match point X on two width of cloth images (x, y, z, l)
TAnd X ' (x ', y ', z ' l)
TSatisfy following relation: X '=HX, wherein H is for describing the transformation matrix of image motion relation, and concrete form is:
The x of match point, the y coordinate can obtain from image, and the z component needs to ask for by other subsidiary condition, and in general big cost of difficulty is higher.Scene around the rotation of camera intrinsic point of fixity is taken, when perhaps the shape of subject was the plane, match point coordinate and projection matrix all can be done corresponding simplification, and the match point coordinate no longer needs the z component, and projection matrix becomes following form:
Be called the homography matrix (homography) of describing object of which movement relation on the plane.The corn ear outside surface is cylindrical, because twice measurement in the front and back anglec of rotation at interval is less, lap is crooked less in the space in two two field pictures of front and back, can be approximated to be the plane motion situation, and therefore, its kinematic relation can be described by homography matrix.
In general, corresponding match point is by the feature description of image, and characteristics of image can be detected by the gray-scale statistical amount, for example, (sum of squared ofdifference, SSD), but this method is subjected to the influence of illumination and geometry deformation to the quadratic sum of calculating pixel point gray value differences easily.Present embodiment is generalized on the image multiscale space feature detection to overcome the influence of illumination and slight geometry deformation, and its yardstick unchangeability is advanced gaussian filtering (x, the y of different scale by the image to different levels, k σ), obtain the Gaussian image L (x, y, k σ) of one group of different scale.Organize the first frame figure based on this, 1/2 ratio carries out the Gaussian image that Filtering Processing obtains second group of different scale to up-sampling to this sub-sampling image, repeats the gaussian pyramid that aforesaid operations obtains image.On the same group two adjacent Gaussian image do difference obtain the difference of Gaussian image (difference of gaussian, DOG),
Peak point be stable characteristics the most, so the peak point on the DOG image is exactly a unique point to be detected.For example, in 8 consecutive point and two-layer up and down each 9 pixel, be extreme value as if the pixel on certain layer of DOG image, then this point is the candidate feature point.The candidate feature point is carried out the quadratic interpolation of 3d space and accurately locate, by (σ) Taylor expansion is removed the less unique point of contrast, calculates the influence that Hessian matrix trace and determinant are removed noise and marginal point for x, y to D.Unique point must be given an anti-rotational direction, in the scalogram picture at unique point place, calculate the neighborhood histogram, surpassing 80% gradient direction with accumulative total is principal direction, and carries out quadratic interpolation in three positions adjacent with principal direction and remove noise effect and finally determine the unique point direction.Detected unique point needs corresponding feature description word (descriptor) and just can mate, similar with the unique point direction, describing word (descriptor) also is to be based upon scalogram as on the gradient orientation histogram in the neighborhood, and has done respective handling for anti-border and illumination effect.Adopt 16x16 neighborhood Gauss's weighting histogram of gradients to add up, thereby obtain the describing word of 4x4x8=128 dimension.Set up kd tree with the describing word of unique point, and take approximate proximity search algorithm that unique point is mated.Can obtain the corresponding match point of many groups by said method.
First search algorithm can produce mistake and mate phenomenon, calculates the projection transition matrix as these corresponding match points of direct use, certainly will influence solving precision and splicing effect.Present embodiment adopts the random sampling consistency algorithm to remove the mistake coupling exterior point (outlier) of match point centering.This algorithm specifically describes as follows: among the conceptual data collection N, the MDS minimum data set m that satisfies model parameter estimation is randomly drawed in circulation; Supposing to have in N the data I match point is correct (inlier), and its ratio is ε=I/N, when N＞＞probability that can carry out once correct sampling during m is ε
mThis algorithm can guarantee can obtain once correct sampling at least in L back of circulation under degree of confidence p, thereby is able to correct estimation model parameter.Cycle index L can be calculated by formula (3):
Owing to have 8 unknown quantitys to determine in the homography matrix, 4 pairs of match points can be found the solution (x in theory, y two components), in 0.99 time supposition of degree of confidence 50% exterior point (the exterior point ratio is much smaller than 50% in the experiment) is arranged, the cycle index that is obtained once correct sampling by following formula as can be known is 72.Concrete steps are as follows:
S1031: from concentrated 4 pairs of match points, the calculating homography matrix H of selecting at random of match point;
S1032: concentrate remaining point to (X to match point
i, X '
i) in X
iApplication H matrix is done conversion and is got X
i h, and calculate X
i hWith X '
iBetween Euclidean distance d (X
i h, X '
i), if d is less than pre-set threshold t (being preferably 4pixel), then match point is to (X
i, X '
i) be the consistent point of sampling of selected 4 pairs of match points among the step S1032;
S1033: if the sampling that calculates by step S1032 consistent put right quantity greater than pre-set threshold T (in the present embodiment, this threshold value T get the candidate matches number of spots 85%), then use consistent put of sampling to recomputate homography matrix, the calculating end; Consistently if sample put right quantity, then reselect point set, estimation model parameter and repeat above-mentioned deterministic process less than T;
S1034: when frequency in sampling surpasses in advance calculated value (72 times), select the sampling consistance point set of quantity maximum to calculate homography matrix.
Multiple image can be registered in the same coordinate system according to the image geometry motion model.Can carry out perspective transform to the former frame image after calculating homography matrix, the former frame image registration is arrived in the coordinate system of back one two field picture, set up one with big new images such as former frame image, the pixel average of two two field picture correspondence positions before and after the new images pixel value is got.
S104: image co-registration; In the new images that described step S103 generates, seek suture line,, make up the section of two two field pictures and form the part panorama sketch according to two two field pictures before and after the cutting of suture line position; Near the suture line position pixel is resampled to eliminate two two field picture difference in exposure about suture line; Particularly, in new images, two original image overlapping region pixels change little, and grey value difference is less, and the pixel value of lap is not subjected to the influence of two original images the factitious transition of gray-scale value to occur, i.e. ghost phenomenon (ghosting).The best suture line of present embodiment employing dynamic programming algorithm searching front and back two two field pictures carries out image slices and comes the removal of images ghost.That is, partly cut out front and back two two field pictures according to picture registration, get the left cut sheet of former frame image, the right cut sheet of back one two field picture is spliced into new images.This dynamic programming algorithm is implemented as follows:
S1041: at first the former frame image after the projective transformation generates the suture path search graph with back one two field picture: establish former frame image x
iThe grey scale pixel value a of place, back one two field picture x '
iThe grey scale pixel value b of place, the suture path search graph
The pixel value c at place can be calculated by formula (4):
By formula (4) as can be known, the grey scale pixel value of suture path search graph in the overlapping region of front and back two two field pictures is less, and other area grayscale values are bigger;
S1042: from first row of above-mentioned suture path search graph, the starting point of each row is designated as a sutural end points;
S1043: begin to search for the next line 3 pixels of adjacency from each sutural end points, the pixel of getting the gray-scale value minimum is as this sutural expansion direction;
S1044: search for downwards up to the image bottom line by line along sutural expansion direction, select a best suture line of conduct of gray-scale value sum minimum in all suture lines.
Sutural position and form in the present embodiment have been shown among Fig. 3.The intermediate section boundary line is suture line.
In the specific implementation process, stitching image is derived from two width of cloth different images, and the variation of light causes the brightness difference of respective pixel during photographic images, so the suture line both sides have certain light and shade difference.The illumination that controls environment can reduce the part difference in exposure, but because body surface optical reflection attribute difference, even strict control illumination condition, the conversion of video camera shooting angle still can make image produce the light and shade difference, add homography matrix and find the solution the influence of error, the pixel of suture line both sides must exist color and locational trickle discrepancy with truth, and this will influence the panorama sketch observing effect, must remove.
Present embodiment is realized the elimination of difference in exposure by the laplacian pyramid blending algorithm of image.This algorithm is to carry out on different scale, different spatial resolutions and the different levels at image, does difference after Gaussian image that its each tomographic image is this layer and last layer image interpolation are amplified and obtaining; Its implementation is a bandpass filtering process, can eliminate difference in exposure under the situation that keeps some image details.Its specific algorithm is as follows:
1, set up the image gaussian pyramid: with the bottom layer image of former figure as gaussian pyramid, and to bottom layer image carry out gaussian filtering and interlacing every row obtain the ground floor image to down-sampling, repeat this operation and set up the image gaussian pyramid;
2,, to the top layer images interpolate value its size is equated with last layer, and do difference acquisition laplacian image pyramid with the last layer image from Gaussian image pyramid top layer;
3, recover image according to formula (5) in the gaussian pyramid mode, obtain the fused images after the Laplace transform:
Wherein G is the image gaussian pyramid, and L is a laplacian pyramid, G
*Be Gaussian image interpolation enlarged image.
The laplacian image conversion can be good at the different light and shade differences that cause of removal of images exposure, but in setting up the process of image pyramid image is carried out operations such as gaussian filtering, unscented transformation, interpolate value the details of image is thickened, blurred picture all has certain influence to observing effect and subsequent treatment.The laplacian image conversion is that the entire image zone is handled, and in fact the difference of suture line both sides pixel value is the most obvious, if can make near the pixel smooth transition the suture line, keep simultaneously apart from suture line than image pixel value is constant at a distance, under difference in exposure was not very big situation, the details and the sharpening degree that can keep image can be eliminated the influence of difference in exposure to observing effect again like this.The present invention is that the boundary sets up the black and white template image with the suture line, and template is carried out gaussian filtering, is that weights treat that from two frames fused images is set up in sampling the fused images with the filtering image pixel value, and sampling template method for building up is as follows:
1, setting up the bianry image of a former figure size, is the boundary with the suture line, and it is 255 that its left pixel is composed, and it is 0 that right pixel is composed.
2, use the 7x7 template to carry out repeatedly gaussian filtering, filter times is decided on the remarkable situation of difference in exposure, employed two-dimensional discrete gaussian filtering template as shown in the formula:
1 4 7 10 7 4 1
4 12 26 33 26 12 4
7 26 55 71 55 26 7
10 33 71 91 71 33 10 (6)
7 26 55 71 55 26 7
4 12 26 33 26 12 4
1 4 7 10 7 4 1
3, the pixel value of establishing a certain position of filtering rear pattern plate image is g, and g is the sampling weights of former frame image, and 255-g is the sampling weights of back one two field picture.
If sampling composograph x
iThe position pixel value is c, the former frame image
The position pixel value is b, back one two field picture
The position pixel is a, template image
The position pixel is g, and then c can be calculated by formula (7):
Light and shade changes the influence that near violent zone (suture line) is subjected to gaussian filtering, and the light and shade variation tendency relaxes, and the area grayscale value far away apart from suture line is close, is subjected to the influence of gaussian filtering limited, can keep the details of original image during sampling as far as possible.Suture line both sides pixel light and shade transition phenomenon disappears substantially, and the difference in exposure of fused images has not influenced observing effect.
In the specific implementation process, if front and back two two field picture difference in exposure are very remarkable, then adopting Laplce's fusion method, is cost unified fusion brightness of image with the sacrificial section image resolution ratio; If difference in exposure is little, then adopt the weights fusion method, when keeping image detail, make suture line both sides image transition mild.
Fig. 4 shows the image mosaic mode of present embodiment, among the figure 1, the 2...n part panorama sketch of representing the n time splicing to obtain respectively.Wherein, because left side section is obtained by the first two field picture perspective transform among the part panoramic picture I ', repeatedly the perspective transform operation accumulated of splicing will make that picture shape diminishes, resolution reduces, and be not easy to observe.Therefore all need after each splicing the image in suture line left side is carried out level and stretched vertically, and its pixel is carried out bilinear interpolation, and pixel moves formed cavity in the drawing process to fill up.
S105: resistance to deformation is handled; Specifically, image carried out perspective projection transformation after, image pixel moves occurred level and vertical direction.Therefore, can after at every turn to image perspective transform, registration, splicing, carry out stretched operation again to offset the distortion of image level and vertical two directions.Stretching side-play amount computing method are shown below:
Suppose spliced figure is carried out stretching on the horizontal direction, wherein d ' is this row pixel maximum offset (can be tried to achieve by homography matrix) on the draw direction, x
eBe this row pixel boundary coordinate after the projective transformation, x
sBe the coordinate of suture line at this row, x is for desiring the stretch position pixel coordinate.The suture line left-side images is stretched according to the stretching side-play amount that formula (8) calculates, and the displacement of vertical direction is corrected by handling with quadrat method.
For fear of the discontinuous generation cavity that pixel moves, set up a blank image, the image that do not stretch is directly duplicated on the suture line right side, the left pixel value is searched in former figure by the result of calculation of pixel coordinate and stretching side-play amount, its coordinate figure may not be an integer, it is carried out bilinear interpolation obtain pixel value.Both eliminate the interference of being out of shape to a certain extent through the image of straightening, and made the image in zone to be matched keep original form constant again, simplified the complexity of images match, strengthened the stability of algorithm.
S106: the image sequence to continuous shooting carries out image registration and fusion treatment according to abovementioned steps S101-105, generates the fruit ear panorama sketch; Specifically:
S1061: difference fruit ear image and background image obtain only to comprise the image sequence of fruit ear target, and carry out gray processing and handle;
The unique point of first frame and second two field picture in S1062: use Harris, SUSAN or the SIFT method extraction fruit ear target binary image sequence, and press gradient direction statistic coupling character pair point;
S1063: find the solution the projective transformation matrix P of first two field picture to second two field picture by least square method, set up a new images I identical with first two field picture size, the pixel of x position can be by position x '=P in first two field picture among the I
-1The x pixel determines that x ' is not integer usually, can pass through the bilinear interpolation approximate evaluation; There are the gray scale similar area in the new images I and second two field picture, in this zone, seek suture line, by the suture line cutting new images (by obtaining after the first two field picture projective transformation) and second two field picture, part panoramic picture I ' is merged in section of new images left side and the section of the second two field picture right side, and suture line both sides pixel resamples by Gauss's template and eliminates the luminance difference of two splicing sections;
S1064: replace second two field picture and the 3rd two field picture to carry out concatenation with part panoramic picture I ', so repeatedly up to handling whole image sequence.
Be the corn ear panorama sketch that generates according to embodiment as shown in Figure 5, wherein, (a) being a frame fruit ear image in the image sequence, (b) is that the splicing of front and back two two field pictures is merged, (c) the fruit ear panorama sketch for not carrying out stretch processing (d) is the fruit ear panorama sketch after the stretch processing.
In sum, the present invention uses the digital camera of fixed position to corn ear interval certain angle photographic images sequence; Extract image characteristic point and, it is registered to the same coordinate system forms a width of cloth new images according to the kinematic parameter of front and back two two field pictures in the matching characteristic point sequence of computed images; In new images, seek suture line, and according to two two field pictures before and after the cutting of suture line position; Suture line both sides pixel exposure difference is eliminated in the filtering of Gauss's template, in level and vertical direction stretching image counteract perspective projection deformation; Image sequence is carried out the synthetic corn ear panorama sketch of above operation successively.Adopt method of the present invention can show the seamless spliced overall picture of corn ear image sequence on piece image, thereby reduced the redundance of photographic intelligence greatly, reduce the information stores space, morphologic observation is more directly perceived.
Above embodiment only is used to illustrate the present invention; and be not limitation of the present invention; the those of ordinary skill in relevant technologies field; under the situation that does not break away from the spirit and scope of the present invention; can also make various variations and modification; therefore all technical schemes that are equal to also belong to category of the present invention, and scope of patent protection of the present invention should be defined by the claims.
Claims (8)
1. a method of using image sequence to generate the corn ear panorama sketch is characterized in that, may further comprise the steps:
S1: obtain the image sequence that a plurality of images of corn ear are formed by the rotation corn and in the mode that at interval predetermined angle is taken continuously, wherein per two frame adjacent images have the overlapping region;
S2: extract the target image sequence of the fruit ear in the described image sequence by the background subtraction point-score, and it is carried out gray processing handle, obtain the target grayscale image sequence of fruit ear;
S3: by perspective projection transformation the 1st frame in the target gray level image sequence of described fruit ear and the 2nd two field picture are registered to the same coordinate system and form a frame new images I;
S4: search for the best suture line of described new images I by dynamic programming algorithm, cut described the 1st frame and the 2nd two field picture, the right side section of the left side section of described the 1st two field picture and the 2nd two field picture is combined as part panoramic picture I ' along described best suture line;
S5: the image that is positioned at described best suture line left side among the described part panoramic picture I ' is carried out level and stretched vertically,, finish once splicing to eliminate the distortion of described part panoramic picture I ' in the projective transformation process;
S6: to the 3rd two field picture execution in step S3-S5 in the target gray level image sequence of part panoramic picture I ' after the stretch processing and described fruit ear to splice next time, so be repeated to the last frame image in the target gray level image sequence of described fruit ear, obtain the corn ear panorama sketch.
2. use image sequence as claimed in claim 1 generates the method for corn ear panorama sketch, it is characterized in that described step S3 further comprises following substep:
S301: extract the unique point in the overlapping region of described the 1st frame and the 2nd two field picture by harris, SUSAN or SIFT method;
S302: according to proximity search algorithm described unique point is mated, obtain the many groups of corresponding match point X and the X ' of two two field pictures;
S303: calculate the described homography matrix H that organizes between corresponding match point X and the X ' by least square method more, wherein, X '=HX;
S304: according to described homography matrix H the 1st two field picture is carried out perspective projection transformation, it is registered to the 2nd two field picture place coordinate system, form new images I;
Wherein, the pixel of described new images I is the average of the 1st two field picture and the 2nd two field picture respective pixel after the projective transformation.
3. use image sequence as claimed in claim 2 generates the method for corn ear panorama sketch, it is characterized in that, also comprises between described step S302 and the S303: remove the described mistake match point of organizing in the corresponding match point by the random sampling consistency algorithm more.
4. use image sequence as claimed in claim 1 generates the method for corn ear panorama sketch, it is characterized in that, among the described step S4, seeks best sutural process by dynamic programming and further comprises:
S401: the first row pixel by described new images I begins, and each is listed as first pixel and is designated as a sutural end points;
S402: begin to search for the next line pixel 3 pixels of adjacency from each sutural end points, the pixel of getting the gray-scale value minimum is as described sutural bearing of trend, and search downwards is until the bottom of described new images I line by line;
S403: repeat above-mentioned steps and obtain many suture lines, select a wherein best suture line of conduct of grey scale pixel value sum minimum.
5. use image sequence as claimed in claim 1 generates the method for corn ear panorama sketch, it is characterized in that, among the described step S4, before cutting described the 1st frame and the 2nd two field picture along described best suture line, also eliminate the difference in exposure of described best suture line both sides image by laplacian pyramid blending algorithm or weights blending algorithm.
6. use image sequence as claimed in claim 1 generates the method for corn ear panorama sketch, it is characterized in that among the described step S5, the pixel-shift amount d of described level and stretched vertically follows following formula:
Wherein d ' is the maximum offset of this row pixel on the draw direction, x
eBe this row pixel boundary coordinate after the projective transformation, x
sBe the coordinate of best suture line at this row, x is for desiring the pixel coordinate of stretch position.
7. use image sequence as claimed in claim 1 generates the method for corn ear panorama sketch, it is characterized in that, among the step S5, after eliminating the distortion of described part panoramic picture I ' in the projective transformation process, also pixel moves the cavity that is produced in the drawing process to fill up to carrying out bilinear interpolation through the pixel of the described part panoramic picture I ' of level and stretched vertically.
8. generate the method for corn ear panorama sketch as any described use image sequence among the claim 1-7, it is characterized in that among the described step S1, the area of overlapping region is greater than 50% of single-frame images area between described adjacent two two field pictures.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201010176315A CN101853524A (en) | 2010-05-13 | 2010-05-13 | Method for generating corn ear panoramic image by using image sequence |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201010176315A CN101853524A (en) | 2010-05-13 | 2010-05-13 | Method for generating corn ear panoramic image by using image sequence |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN101853524A true CN101853524A (en) | 2010-10-06 |
Family
ID=42804989
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201010176315A Pending CN101853524A (en) | 2010-05-13 | 2010-05-13 | Method for generating corn ear panoramic image by using image sequence |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN101853524A (en) |
Cited By (42)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102096915A (en) * | 2011-02-09 | 2011-06-15 | 北京航空航天大学 | Camera lens cleaning method based on precise image splicing |
| CN102164269A (en) * | 2011-01-21 | 2011-08-24 | 北京中星微电子有限公司 | Method and device for monitoring panoramic view |
| WO2012089116A1 (en) * | 2010-12-27 | 2012-07-05 | 汉王科技股份有限公司 | Device and method for scanning and recognizing |
| CN102982524A (en) * | 2012-12-25 | 2013-03-20 | 北京农业信息技术研究中心 | Splicing method for corn ear order images |
| CN103020970A (en) * | 2012-12-25 | 2013-04-03 | 北京农业信息技术研究中心 | Corn ear image grain segmentation method |
| CN103024263A (en) * | 2011-09-26 | 2013-04-03 | 卡西欧计算机株式会社 | Image processing device and image processing method |
| CN103020941A (en) * | 2012-12-28 | 2013-04-03 | 昆山市工业技术研究院有限责任公司 | Panoramic stitching based rotary camera background establishment method and panoramic stitching based moving object detection method |
| CN103093456A (en) * | 2012-12-25 | 2013-05-08 | 北京农业信息技术研究中心 | Corn ear character index computing method based on images |
| CN103208072A (en) * | 2012-01-14 | 2013-07-17 | 梁洛文 | Furniture matching showing method |
| CN103501415A (en) * | 2013-10-01 | 2014-01-08 | 中国人民解放军国防科学技术大学 | Overlap structural deformation-based video real-time stitching method |
| CN103501409A (en) * | 2013-10-01 | 2014-01-08 | 中国人民解放军国防科学技术大学 | Ultrahigh resolution panorama speed dome AIO (All-In-One) system |
| US20140023243A1 (en) * | 2012-07-23 | 2014-01-23 | Dow Agrosciences Llc | Kernel counter |
| CN104349025A (en) * | 2013-08-09 | 2015-02-11 | 联想(北京)有限公司 | Panoramic picture generation method and device as well as electronic equipment |
| CN104680501A (en) * | 2013-12-03 | 2015-06-03 | 华为技术有限公司 | Image splicing method and device |
| CN104865259A (en) * | 2015-05-07 | 2015-08-26 | 中国农业大学 | Falling type corn ear holographic character rapid measuring system and method |
| CN104933687A (en) * | 2015-07-09 | 2015-09-23 | 武汉大学 | Seam line multiscale feather algorithm of considering changed area |
| CN105574815A (en) * | 2015-12-21 | 2016-05-11 | 湖南优象科技有限公司 | Image splicing method and device used for scanning mouse |
| CN105631811A (en) * | 2016-02-25 | 2016-06-01 | 科盾科技股份有限公司 | Image stitching method and device |
| CN105809815A (en) * | 2015-01-19 | 2016-07-27 | 株式会社巨晶片 | Characteristic image generation device, determination device, characteristic image generation and determination method |
| CN106204456A (en) * | 2016-07-18 | 2016-12-07 | 电子科技大学 | Panoramic video sequences estimation is crossed the border folding searching method |
| CN106204507A (en) * | 2015-05-28 | 2016-12-07 | 长沙维纳斯克信息技术有限公司 | A kind of unmanned plane image split-joint method |
| CN106846385A (en) * | 2016-12-30 | 2017-06-13 | 广州地理研究所 | Many sensing Remote Sensing Images Matching Methods, device and system based on unmanned plane |
| CN106954012A (en) * | 2017-03-29 | 2017-07-14 | 武汉嫦娥医学抗衰机器人股份有限公司 | A kind of high definition polyphaser full-view stereo imaging system and method |
| CN103718540B (en) * | 2011-02-21 | 2017-11-07 | 英特尔公司 | Method, user terminal and computer readable recording medium storing program for performing for generating panoramic picture |
| CN107644411A (en) * | 2017-09-19 | 2018-01-30 | 武汉中旗生物医疗电子有限公司 | Ultrasonic wide-scene imaging method and device |
| CN108362698A (en) * | 2018-02-06 | 2018-08-03 | 北京农业智能装备技术研究中心 | A kind of rice shoot stem nodal test method and device |
| CN109300432A (en) * | 2017-07-24 | 2019-02-01 | 晶门科技（中国）有限公司 | The method that gray level image shows signal driving is carried out in monochromatic display panel |
| CN110310251A (en) * | 2019-07-03 | 2019-10-08 | 北京字节跳动网络技术有限公司 | Image processing method and device |
| CN110349174A (en) * | 2019-06-28 | 2019-10-18 | 佛山科学技术学院 | A kind of sliding rail Multi-parameter Measurement Method and measuring device |
| CN110490868A (en) * | 2019-08-23 | 2019-11-22 | 陈�峰 | A kind of lossless method of counting based on computer vision corn grain number per spike |
| CN110864644A (en) * | 2019-11-13 | 2020-03-06 | 北京农业信息技术研究中心 | Plant stem phenotype information acquisition device and method |
| CN111652916A (en) * | 2020-05-11 | 2020-09-11 | 浙江大华技术股份有限公司 | Panoramic image generation method, panoramic image generation device and computer storage medium |
| CN111709976A (en) * | 2020-08-24 | 2020-09-25 | 湖南国科智瞳科技有限公司 | Rapid registration method and system for microscopic image and computer equipment |
| WO2020228680A1 (en) * | 2019-05-10 | 2020-11-19 | 国网浙江省电力有限公司台州供电公司 | Dual camera image-based splicing method and apparatus, and electronic device |
| CN112529028A (en) * | 2019-09-19 | 2021-03-19 | 北京声迅电子股份有限公司 | Networking access method and device for security check machine image |
| CN112734720A (en) * | 2021-01-08 | 2021-04-30 | 沈阳工业大学 | Ship hull laser cleaning in-place detection method and system based on visual identification |
| CN112783839A (en) * | 2020-06-08 | 2021-05-11 | 北京金山办公软件股份有限公司 | Method and device for storing document, electronic equipment and storage medium |
| US20210287392A1 (en) * | 2020-03-13 | 2021-09-16 | Oregon State University | Novel system to quantify maize seed phenotypes |
| CN113962867A (en) * | 2021-12-22 | 2022-01-21 | 深圳思谋信息科技有限公司 | Image processing method, image processing device, computer equipment and storage medium |
| CN114007014A (en) * | 2021-10-29 | 2022-02-01 | 北京环境特性研究所 | Method and device for generating panoramic image, electronic equipment and storage medium |
| WO2023165452A1 (en) * | 2022-03-04 | 2023-09-07 | 华为技术有限公司 | Motion information acquisition method, calibration method, and apparatus |
| CN117336620A (en) * | 2023-11-24 | 2024-01-02 | 北京智汇云舟科技有限公司 | Adaptive video stitching method and system based on deep learning |
-
2010
- 2010-05-13 CN CN201010176315A patent/CN101853524A/en active Pending
Cited By (65)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2012089116A1 (en) * | 2010-12-27 | 2012-07-05 | 汉王科技股份有限公司 | Device and method for scanning and recognizing |
| US9565358B2 (en) | 2010-12-27 | 2017-02-07 | Hanwang Technology Co., Ltd. | Apparatus and method for scanning and recognizing |
| CN102164269A (en) * | 2011-01-21 | 2011-08-24 | 北京中星微电子有限公司 | Method and device for monitoring panoramic view |
| CN102096915A (en) * | 2011-02-09 | 2011-06-15 | 北京航空航天大学 | Camera lens cleaning method based on precise image splicing |
| CN102096915B (en) * | 2011-02-09 | 2013-08-07 | 北京航空航天大学 | Camera lens cleaning method based on precise image splicing |
| CN103718540B (en) * | 2011-02-21 | 2017-11-07 | 英特尔公司 | Method, user terminal and computer readable recording medium storing program for performing for generating panoramic picture |
| CN103024263A (en) * | 2011-09-26 | 2013-04-03 | 卡西欧计算机株式会社 | Image processing device and image processing method |
| CN103024263B (en) * | 2011-09-26 | 2016-08-03 | 卡西欧计算机株式会社 | Image processing apparatus and image processing method |
| CN103208072A (en) * | 2012-01-14 | 2013-07-17 | 梁洛文 | Furniture matching showing method |
| US20140023243A1 (en) * | 2012-07-23 | 2014-01-23 | Dow Agrosciences Llc | Kernel counter |
| CN103093456A (en) * | 2012-12-25 | 2013-05-08 | 北京农业信息技术研究中心 | Corn ear character index computing method based on images |
| CN103020970A (en) * | 2012-12-25 | 2013-04-03 | 北京农业信息技术研究中心 | Corn ear image grain segmentation method |
| CN103020970B (en) * | 2012-12-25 | 2016-03-30 | 北京农业信息技术研究中心 | Corn ear image grain segmentation method |
| CN102982524A (en) * | 2012-12-25 | 2013-03-20 | 北京农业信息技术研究中心 | Splicing method for corn ear order images |
| CN102982524B (en) * | 2012-12-25 | 2015-03-25 | 北京农业信息技术研究中心 | Splicing method for corn ear order images |
| CN103093456B (en) * | 2012-12-25 | 2015-06-03 | 北京农业信息技术研究中心 | Corn ear character index computing method based on images |
| CN103020941A (en) * | 2012-12-28 | 2013-04-03 | 昆山市工业技术研究院有限责任公司 | Panoramic stitching based rotary camera background establishment method and panoramic stitching based moving object detection method |
| CN104349025B (en) * | 2013-08-09 | 2018-12-14 | 联想(北京)有限公司 | A kind of panorama image generation method, device and electronic equipment |
| CN104349025A (en) * | 2013-08-09 | 2015-02-11 | 联想(北京)有限公司 | Panoramic picture generation method and device as well as electronic equipment |
| CN103501415B (en) * | 2013-10-01 | 2017-01-04 | 中国人民解放军国防科学技术大学 | A kind of real-time joining method of video based on lap malformation |
| CN103501415A (en) * | 2013-10-01 | 2014-01-08 | 中国人民解放军国防科学技术大学 | Overlap structural deformation-based video real-time stitching method |
| CN103501409A (en) * | 2013-10-01 | 2014-01-08 | 中国人民解放军国防科学技术大学 | Ultrahigh resolution panorama speed dome AIO (All-In-One) system |
| CN104680501A (en) * | 2013-12-03 | 2015-06-03 | 华为技术有限公司 | Image splicing method and device |
| CN104680501B (en) * | 2013-12-03 | 2018-12-07 | 华为技术有限公司 | The method and device of image mosaic |
| CN105809815A (en) * | 2015-01-19 | 2016-07-27 | 株式会社巨晶片 | Characteristic image generation device, determination device, characteristic image generation and determination method |
| CN105809815B (en) * | 2015-01-19 | 2019-01-04 | 株式会社巨晶片 | Characteristic image generating means and decision maker and characteristic image generation method and determination method |
| CN104865259A (en) * | 2015-05-07 | 2015-08-26 | 中国农业大学 | Falling type corn ear holographic character rapid measuring system and method |
| CN106204507A (en) * | 2015-05-28 | 2016-12-07 | 长沙维纳斯克信息技术有限公司 | A kind of unmanned plane image split-joint method |
| CN104933687B (en) * | 2015-07-09 | 2018-01-23 | 武汉大学 | A kind of multiple dimensioned emergence algorithm of jointing line for considering region of variation |
| CN104933687A (en) * | 2015-07-09 | 2015-09-23 | 武汉大学 | Seam line multiscale feather algorithm of considering changed area |
| CN105574815A (en) * | 2015-12-21 | 2016-05-11 | 湖南优象科技有限公司 | Image splicing method and device used for scanning mouse |
| CN105631811A (en) * | 2016-02-25 | 2016-06-01 | 科盾科技股份有限公司 | Image stitching method and device |
| CN106204456B (en) * | 2016-07-18 | 2019-07-19 | 电子科技大学 | Panoramic video sequences estimation, which is crossed the border, folds searching method |
| CN106204456A (en) * | 2016-07-18 | 2016-12-07 | 电子科技大学 | Panoramic video sequences estimation is crossed the border folding searching method |
| CN106846385B (en) * | 2016-12-30 | 2020-11-24 | 广州地理研究所 | Multi-sensing remote sensing image matching method, device and system based on unmanned aerial vehicle |
| CN106846385A (en) * | 2016-12-30 | 2017-06-13 | 广州地理研究所 | Many sensing Remote Sensing Images Matching Methods, device and system based on unmanned plane |
| CN106954012A (en) * | 2017-03-29 | 2017-07-14 | 武汉嫦娥医学抗衰机器人股份有限公司 | A kind of high definition polyphaser full-view stereo imaging system and method |
| CN109300432A (en) * | 2017-07-24 | 2019-02-01 | 晶门科技（中国）有限公司 | The method that gray level image shows signal driving is carried out in monochromatic display panel |
| CN107644411A (en) * | 2017-09-19 | 2018-01-30 | 武汉中旗生物医疗电子有限公司 | Ultrasonic wide-scene imaging method and device |
| CN108362698B (en) * | 2018-02-06 | 2020-08-11 | 北京农业智能装备技术研究中心 | Method and device for detecting main stem nodes of seedlings |
| CN108362698A (en) * | 2018-02-06 | 2018-08-03 | 北京农业智能装备技术研究中心 | A kind of rice shoot stem nodal test method and device |
| WO2020228680A1 (en) * | 2019-05-10 | 2020-11-19 | 国网浙江省电力有限公司台州供电公司 | Dual camera image-based splicing method and apparatus, and electronic device |
| CN110349174A (en) * | 2019-06-28 | 2019-10-18 | 佛山科学技术学院 | A kind of sliding rail Multi-parameter Measurement Method and measuring device |
| CN110310251A (en) * | 2019-07-03 | 2019-10-08 | 北京字节跳动网络技术有限公司 | Image processing method and device |
| CN110310251B (en) * | 2019-07-03 | 2021-10-29 | 北京字节跳动网络技术有限公司 | Image processing method and device |
| CN110490868A (en) * | 2019-08-23 | 2019-11-22 | 陈�峰 | A kind of lossless method of counting based on computer vision corn grain number per spike |
| CN110490868B (en) * | 2019-08-23 | 2023-05-30 | 陈�峰 | Nondestructive counting method based on computer vision corn cob grain number |
| CN112529028B (en) * | 2019-09-19 | 2022-12-02 | 北京声迅电子股份有限公司 | Networking access method and device for security check machine image |
| CN112529028A (en) * | 2019-09-19 | 2021-03-19 | 北京声迅电子股份有限公司 | Networking access method and device for security check machine image |
| CN110864644A (en) * | 2019-11-13 | 2020-03-06 | 北京农业信息技术研究中心 | Plant stem phenotype information acquisition device and method |
| US20210287392A1 (en) * | 2020-03-13 | 2021-09-16 | Oregon State University | Novel system to quantify maize seed phenotypes |
| US11823408B2 (en) * | 2020-03-13 | 2023-11-21 | Oregon State University | Apparatus and method to quantify maize seed phenotypes |
| CN111652916B (en) * | 2020-05-11 | 2023-09-29 | 浙江大华技术股份有限公司 | Panoramic image generation method, panoramic image generation device and computer storage medium |
| CN111652916A (en) * | 2020-05-11 | 2020-09-11 | 浙江大华技术股份有限公司 | Panoramic image generation method, panoramic image generation device and computer storage medium |
| CN112783839A (en) * | 2020-06-08 | 2021-05-11 | 北京金山办公软件股份有限公司 | Method and device for storing document, electronic equipment and storage medium |
| CN111709976A (en) * | 2020-08-24 | 2020-09-25 | 湖南国科智瞳科技有限公司 | Rapid registration method and system for microscopic image and computer equipment |
| CN112734720A (en) * | 2021-01-08 | 2021-04-30 | 沈阳工业大学 | Ship hull laser cleaning in-place detection method and system based on visual identification |
| CN112734720B (en) * | 2021-01-08 | 2024-03-05 | 沈阳工业大学 | Ship hull laser cleaning in-situ detection method and system based on visual identification |
| CN114007014A (en) * | 2021-10-29 | 2022-02-01 | 北京环境特性研究所 | Method and device for generating panoramic image, electronic equipment and storage medium |
| CN114007014B (en) * | 2021-10-29 | 2023-06-16 | 北京环境特性研究所 | Method and device for generating panoramic image, electronic equipment and storage medium |
| CN113962867A (en) * | 2021-12-22 | 2022-01-21 | 深圳思谋信息科技有限公司 | Image processing method, image processing device, computer equipment and storage medium |
| CN113962867B (en) * | 2021-12-22 | 2022-03-15 | 深圳思谋信息科技有限公司 | Image processing method, image processing device, computer equipment and storage medium |
| WO2023165452A1 (en) * | 2022-03-04 | 2023-09-07 | 华为技术有限公司 | Motion information acquisition method, calibration method, and apparatus |
| CN117336620A (en) * | 2023-11-24 | 2024-01-02 | 北京智汇云舟科技有限公司 | Adaptive video stitching method and system based on deep learning |
| CN117336620B (en) * | 2023-11-24 | 2024-02-09 | 北京智汇云舟科技有限公司 | Adaptive video stitching method and system based on deep learning |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN101853524A (en) | Method for generating corn ear panoramic image by using image sequence | |
| US20230410337A1 (en) | Method and system for scene image modification | |
| CN105245841B (en) | A kind of panoramic video monitoring system based on CUDA | |
| CN102834845B (en) | The method and apparatus calibrated for many camera heads | |
| US9811946B1 (en) | High resolution (HR) panorama generation without ghosting artifacts using multiple HR images mapped to a low resolution 360-degree image | |
| CN105957007B (en) | Image split-joint method based on characteristic point plane similarity | |
| CN103337094B (en) | A kind of method of applying binocular camera and realizing motion three-dimensional reconstruction | |
| US9214013B2 (en) | Systems and methods for correcting user identified artifacts in light field images | |
| TWI554976B (en) | Surveillance systems and image processing methods thereof | |
| KR101121034B1 (en) | System and method for obtaining camera parameters from multiple images and computer program products thereof | |
| JP4668220B2 (en) | Image processing apparatus, image processing method, and program | |
| CN107346041B (en) | Method and device for determining grating parameters of naked eye 3D display equipment and electronic equipment | |
| CN107346040B (en) | Method and device for determining grating parameters of naked eye 3D display equipment and electronic equipment | |
| CN110288511B (en) | Minimum error splicing method and device based on double camera images and electronic equipment | |
| WO2016141810A1 (en) | Image processing method for shooting terminal and apparatus therefor | |
| CN112801870B (en) | Image splicing method based on grid optimization, splicing system and readable storage medium | |
| US20210065336A1 (en) | Method for generating a reduced-blur digital image | |
| JP6285686B2 (en) | Parallax image generation device | |
| WO2020151078A1 (en) | Three-dimensional reconstruction method and apparatus | |
| JP6799468B2 (en) | Image processing equipment, image processing methods and computer programs | |
| CN110136077B (en) | Multi-view point cloud fusion method based on projection | |
| CN111583388A (en) | Scanning method and device of three-dimensional scanning system | |
| CN107251089B (en) | Image processing method for motion detection and compensation | |
| Abdul-Rahim et al. | An in depth review paper on numerous image mosaicing approaches and techniques | |
| Jiang | Stereo vision for facet type cameras |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C12 | Rejection of a patent application after its publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20101006 |