CN101753771B

CN101753771B - Method for solving panoramic image parameters automatically

Info

Publication number: CN101753771B
Application number: CN2008100441047A
Authority: CN
Inventors: 徐小明
Original assignee: SHANGHAI JIETU SOFTWARE TECHN CO Ltd
Current assignee: SHANGHAI JIETU SOFTWARE TECHN CO Ltd
Priority date: 2008-12-11
Filing date: 2008-12-11
Publication date: 2011-10-05
Anticipated expiration: 2028-12-11
Also published as: CN101753771A

Abstract

The invention discloses a method for solving panoramic image parameters automatically, and the method comprises the following steps: searching for connected components by adopting breadth-first search algorithm, optimizing the initial position relation between the connected components, then putting initial position information into a global function, and optimizing global parameters. The method can reduce the number of initial parameter values which need to be set when optimizing, and simultaneously, the initial parameter values are not required, thus reducing the complexity of setting parameters when optimizing the parameters and increasing the accuracy of the image parameters.

Description

A kind of method of solving panoramic image parameters automatically

Technical field:

The present invention relates to a kind of method for solving of panorama sketch, particularly relate to a kind of method of the picture of many collections being carried out the Automatic Optimal global parameter.

Background technology:

Panorama sketch is that a kind of interactively virtual scene is represented mode, and the mode that it is drawn based on image has been reproduced three-dimensional scenic, and available panorama player is realized the roaming of virtual scene, brings real scene to experience to the user.

A core problem is the parameter information that need find the solution every width of cloth image in the panorama sketch generation technique, and the quality of parametric solution has directly determined the quality of last panorama sketch.Panorama parameter optimization technology before all is a prior given initial value, and then be optimized and obtain global parameter, do the restriction when having brought pictures taken undoubtedly like this, need come pictures taken by certain rule, and choosing last influential effect of initial value is bigger, and this has also increased the processing time in later stage, so how a kind of panorama sketch generation technique that is not subjected to the pictures taken restriction is provided, will be the problem that urgent need will solve.

Summary of the invention:

The present invention is directed to existing defective in the above-mentioned existing panorama generation technique, providing a kind of can carry out the automatic calculation panoramic image parameters to many collection pictures, and obtains the method for panorama sketch according to these parameters.

In order to achieve the above object, the technical solution used in the present invention:

A kind of method of solving panoramic image parameters automatically, this method is applied in the panorama sketch, and it comprises the steps:

(1) utilize camera to take multiple image, and to image respectively the search characteristics point mate, it is right to draw match point;

(2) right according to the match point that obtains in the step (1), the image relation is generated a undirected connected graph;

(3) the undirected connected graph that is obtained by step (2) is searched for the breadth First ergodic algorithm, found the precedence relationship between image,, solve the initial position message of image again according to this relation;

(4) positional information that will be obtained by step (3) solves the global parameter of image as initial value, according to these parameters, picture is projected in the panorama, generates panoramic picture.

Take multiple image in the described step (1), coplane is vowed at the pairing center of its horizontal image, and should have certain coincidence zone between the adjacent image.

Take multiple image in the described step (1), center of its top figure or base map is vowed vertical with the centered sagittal plane of horizontal image, and with horizontal image between should have certain zone that overlaps.

To image search characteristics point and when mating respectively, can carry out not according to the order of image in the described step (1), adopt the kd_tree global search, it is right to obtain characteristic matching point.

Described step (2) is when specifically implementing, and is right according to the match point that step (1) obtains, and the image that will be complementary couples together with a limit, and the sequence number of image is the summit sign, generates a undirected connected graph.

Described step (3) is searched for the breadth First ergodic algorithm undirected connected graph, find the precedence relationship between image, again according to this relation, the initial position message that solves image is meant: select a summit arbitrarily, search for the breadth First ergodic algorithm, generate a shortest path that comprises all summits, determine precedence relationship between image with this path; According to this precedence relationship, obtain the positional information of two width of cloth images that are complementary earlier by the LM iterative optimization method again, and to select first width of cloth picture be the reference position, other image is by the relative value adjustment, initial position message to the end.

Described step (4) with the positional information obtained as initial value, the global parameter that solves image is meant: use the binding adjustment algorithm to be optimized as initial value to global parameter the initial position value that every width of cloth picture is obtained, the parameter of global optimization is: lens distortion and horizontal view angle parameter.

The present invention who obtains according to technique scheme carries out the Automatic Optimal global parameter to the picture of many collections, can reduce simultaneously the number that needs the setup parameter initial value when optimizing, and the initial value to parameter does not require yet, the complexity that parameter is provided with when having avoided parameter optimization like this, and improved the accuracy of image parameter.

Description of drawings:

Further specify the present invention below in conjunction with the drawings and specific embodiments.

Fig. 1 is a sample image involved in the present invention.

Fig. 2 is a flow chart of the present invention.

Embodiment:

For technological means, creation characteristic that the present invention is realized, reach purpose and effect is easy to understand, below in conjunction with concrete diagram, further set forth the present invention.

Referring to Fig. 1, it has provided the sample image among the present invention.By the soft-shelled turtle eye pattern picture that fish eye lens is taken, shown in figure A, the fish eye images of the type is used more extensive, but the CCD utilance is lower, and the CCD of the black region beyond the border circular areas is not used to.

Therefore, in order to improve the CCD utilance, camera lens manufacturer has released the fish eye images of sigma camera lens and fullframe camera lens, and the image that the both sides shown in figure B are cut is a drum shaft images, and what four limits shown in the figure C were all cut is full frame image.Shown in the figure D is the common lens image, and figure E is the wide-angle lens imaging.

Method provided by the invention, for many common, wide-angles, soft-shelled turtle eye, drum shaft images, full frame image, can both the Automatic Optimal image parameter.Concrete enforcement can be referring to Fig. 2:

Step 1: at first, the user obtains many images by shooting, makes the center of many images that horizontal direction is taken vow (being the optical axis direction of camera gun) coplane simultaneously, and should there be overlapping scope in adjacent two photos.For drum shaft images and full frame image, also can take the top and bottom image in accordance with regulations, promptly center of its top figure or base map vow vertical with the centered sagittal plane of horizontal image, and with horizontal image between should exist overlap regional; This image can be used to inlay with horizontal image, to generate complete panoramic picture.

Step 2: captured image is carried out characteristic point search and coupling.The characteristic point of all pictures is generated an overall chained list, then each element in the chained list being carried out abutment points with the kd_tree algorithm searches, the generation abutment points is right, and with abutment points to putting into corresponding picture matching relationship, such as abutment points (1,2) is just put in the images match (1,2), the sequence number of 1,2 top presentation video.At last, if exist in the images match relation more than three pairs abutment points to the time, just adopt the RANSAC algorithm to filter, it is right to obtain final characteristic matching point.

For horizontal image, and carry out Feature Points Matching between the adjacent image; For shooting the image at top, the end is arranged, the top and bottom image is mated with every horizontal image respectively; For the soft-shelled turtle eye pattern picture that has only two top and bottom images, then will mate between top image and the base map picture; When mating in this step, can carry out not according to the order of image, it is right to obtain characteristic matching point.

Step 3: the characteristic matching point between the image that obtains according to step 2 is right, set up undirected connected graph and describe relation between the image, concrete grammar is to be the summit of figure with the picture sequence numbers, if matching relationship is arranged between the picture, just inserts on one side between two pairing summits of picture.

Step 4: to the undirected connected graph that generates, select a summit arbitrarily, travel through, generate a shortest path that comprises all summits, determine precedence relationship between image with this path with the breadth First ergodic algorithm;

Step 5: according to this precedence relationship, obtain the positional information (yaw of two width of cloth images that are complementary earlier by the LM iterative optimization method, pitch, roll), and to select first width of cloth picture be the reference position, and the positional information (yaw=0 of first width of cloth picture is set, pitch=0, roll=0), the positional information of other image is adjusted by the adjusting range of reference position, gets initial position message to the end.

Step 6:, camera lens is carried out the correction and the horizontal view angle Parameter Optimization of distortion parameter with the binding adjustment algorithm at last with the initial value of initial position as global optimization.Parameter according to after these optimizations projects to picture in the panorama, generates panoramic picture.

This method that obtains according to technique scheme utilizes the BFS algorithm to seek the assembly that is connected earlier, and the initial position relation of optimization between them then, put into overall situation function with initial positional information, carries out the parameter optimization of the overall situation.

More than show and described basic principle of the present invention and principal character and advantage of the present invention.The technical staff of the industry should understand; the present invention is not restricted to the described embodiments; that describes in the foregoing description and the specification just illustrates principle of the present invention; without departing from the spirit and scope of the present invention; the present invention also has various changes and modifications, and these changes and improvements all fall in the claimed scope of the invention.The claimed scope of the present invention is defined by appending claims and equivalent thereof.

Claims

1. the method for a solving panoramic image parameters automatically, this method is applied to it is characterized in that in the panorama sketch that this method comprises the steps:

(1) utilize camera to take many common, wide-angles, soft-shelled turtle eye, drum shaft images or full frame image, coplane is vowed at the pairing center of the image that its horizontal direction is taken, and should have certain coincidence zone between the adjacent image, and to image respectively the search characteristics point mate, it is right to draw match point;

(2) right according to the match point that obtains in the step (1), be the summit of figure with the picture numbers, if matching relationship is arranged between the image, just between two pairing summits of image, insert on one side, the image relation is generated a undirected connected graph;

(3) the undirected connected graph to being obtained by step (2) is selected a summit arbitrarily, travels through with the breadth First ergodic algorithm, generates a shortest path that comprises all summits, determines precedence relationship between image with this path; Again according to this precedence relationship, obtain the positional information of two width of cloth images that are complementary earlier by the LM iterative optimization method, and to select first width of cloth picture be the reference position, the positional information of first width of cloth picture is set, the positional information of other image is adjusted by the adjusting range of reference position, gets initial position message to the end;

(4) initial position that will be obtained by step (3) is as the initial value of global optimization, with the binding adjustment algorithm camera lens carried out the correction and the horizontal view angle Parameter Optimization of distortion parameter, according to these parameters, picture projected in the panorama, generates panoramic picture.

2. the method for a kind of solving panoramic image parameters automatically according to claim 1, it is characterized in that, in the described step (1) for drum shaft images and full frame image, take the top and bottom image in accordance with regulations, be that the center of its top image or base map picture vows vertical with the centered sagittal plane of horizontal image, and with horizontal image between should exist overlap regional; The top and bottom image can be used to inlay with horizontal image, to generate complete panoramic picture.

3. the method for a kind of solving panoramic image parameters automatically according to claim 1, it is characterized in that, to image search characteristics point and when mating respectively, can carry out in the described step (1) not according to the order of image, adopt the kd_tree global search, it is right to obtain characteristic matching point.