Camera ceramic tile is utilized to realize real-time multiple dimensioned imaging system and method
Technical field
The present invention relates to the use of camera " ceramic tile " and realize real-time multiple dimensioned imaging system and method, belong to image procossing skill
Art field.
Background technology
Digital image resolution depends on pixel count.The resolution of solid-state imaging technology from start 1,000,000 till now 10
To 100,000,000.The least rank to several optical wavelength of current sensor pixel size.Under this yardstick, reduce further
Pixel Dimensions becomes highly difficult.So in order to increase image size, starting to increase sensor area.But at present more than 1 square
Centimetre sensor expensive, output capacity is the highest.It addition, the resolvable spot size of camera lens and space frequency response ability are along with sensing
The increase of device size and reduce.This is also by utilizing camera array rather than should to promote one camera sensor resolution, carries
Rise the reason of image size.
In the last hundred years, panoramic high-definition image is all by rotating one camera on spider, be continuously shot plurality of pictures, spelling
It is connected into a big image.Before digital Age, panoramic picture is to be spliced by shearing.The splicing occurred recently leads to
Cross and realize splicing at association overlapping region common ground.But the most often make a mistake, cause perceptible error.
Main cause is that different pictures shoots at different time, and common ground may there occurs movement between double exposure.
As it is shown in figure 1, panoramic picture can also go to gather by camera array.Camera array can with synchronous acquisition,
Eliminate the time difference between image.United States Patent (USP) 8,259,212 (multiple dimensioned optical system) have been described with this super
High Resolution Image Data Acquisition System.In this framework, it is ensured that it is relatively difficult for having the optical axis of camera to try one's best by same point more
's.
As in figure 2 it is shown, the angle range of camera looks into fee scene is called the angle of visual field.Each camera pixel gathers number from scene
The angle range corresponding according to, pixel is called transient state angle of visual field ifov.One camera has N number of horizontal pixel, M vertical pixel,
Horizontal field of view angle is N*ifov, and vertical field of view angle is M*ifov.Total pixel number is N*M.Can be by that as described in accompanying drawing
Sample utilizes multiple camera or sensor to increase the angle of visual field.Part weight is had between a part of scene of each camera looks into fee and image
Folded, scene can not have the place not covered by camera.
As described in Fig. 2, camera different piece from different view scenes.The different attitudes of camera make
A single visual angle must be formed and always scheme more difficulty.Ideally, it is intended that by camera entrance pupil around a center
Point is arranged, in order to all cameras share a visual angle central point.In practice, this layout is difficult to, and even may not realize.As
Alternative method, we can guarantee that camera optical axis passes through same point, as shown in Figure 3 simply.
In this case, we can move to all camera altogether effective visual angle from aperture entrance by zoomed image
With the visual angle point intersected.But, the amount of zoom that each object is carried out to translate effective visual angle point is the distance phase with object
Close.In the scene of an object with multiple different location, by scaling accurately, to realize translation be impossible.
In practice before, the image of multiple cameras is by being spliced to form a width panorama mosaic figure.Owing to having between camera
Overlapping region, some pixels are by more than one collected by camera, so the total pixel number of composite diagram is less than the pixel of camera matrix
Number.But the pixel count of the most remote super each camera of the pixel count of composite diagram, the farthest surpasses and can the most effectively show
Pixel count.These big figures are generally stored in internal memory, as shown in Figure 4 with " ceramic tile (tile) " form.The figure that resolution is minimum
As in the pyramidal the superiors of ceramic tile, showing scene overall picture.In this figure, low-resolution image is 128*128 pixel.Pyramid
The second layer have 4 width images, every width figure to be 128*128 pixel, scene is divided into four quadrants, increase resolution twice.Next
Layer is made up of 16 tile, is also 128*128 pixel, than the tile increase resolution twice of the second layer.Show at image
In system, the tile that only there is a need to display is used.Such as, the display of a 128*128 only uses the ceramic tiles of the superiors
Sheet shows panorama sketch.When user scales a details of scape of marching into the arena, and only relevant with this details tile is sent to display
Device.In practice before, this tile pyramid structure is that the full resolution figure from synthesis scene starts to create.This knot
Structure makes user can need not see full resolution image at transmission over networks full resolution image.The method is by Baidu, paddy
Song map uses, and is allowing user browse great amount of images data set without in the case of transmitting total data.
The key components invented as this, it was noted that create to use the high-definition image in camera matrix
Global image there is no need in fact, is not even meddlesome.User just hopes and the image from multiple cameras is made us with visual perception
The effect of shock is integrated on display.The splicing mistake of camera adjacent area can cause the dislike of user, from a camera
Visual angle moves to next camera perspective and the most not necessarily causes user to dislike.From this viewpoint, we seek one and do not produce
Raw stitching error, the method translating camera perspective between adjacent cameras.
Under specific circumstances, due to different cameral it can be seen that the scene that can't see of other camera being blocked, so from
It is impossible for calculating angle and eliminating stitching error.In other cases, eliminate by coupling adjacent cameras image or subtract
Few stitching error is possible, but required amount of calculation is also the biggest.In this invention, we seek a kind of method, no
Only avoid stitching error, moreover it is possible to the step for of by eliminating stitching image, ensure displaying live view in the scene.
Recognize that the resolution of single-sensor camera or camera array is far more than display resolution and understands that this invents
Crucial.The resolution of existing HDTV image is 1920*1080 pixel, or 4k (3840*2160 pixel).8k resolution
The camera of (doubling existing 4k resolution camera) is in the middle of exploitation.Recognize high definition resolution for most people display should
With the most enough, additional camera resolution can be used to realize numeral PTZ effect.As described below, with this
Mode uses camera can realize the real time panoramic image without stitching error.
Summary of the invention
The purpose of the present invention: aim to provide one and can reduce or avoid to use image mosaic technology, it is to avoid splicing is by mistake
Difference, promotes panoramic picture display effect, and ensure in panoramic picture scene the step for of by eliminating stitching image real-time
Browse.
This camera " ceramic tile " is utilized to realize real-time multiple dimensioned imaging system, including camera array, image procossing and depositing
Send to display after the image procossing that camera array is obtained by storage system, display, described image procossing and storage system;Its
Being characterised by: described camera array is made up of the microfacies machine of some interlaced settings, described microfacies machine shooting the micro-of gained regard
The resolution of field is higher than the resolution of the display visual field of display.
Described camera array also includes a wide visual field camera.
One quadrant weight of micro-visual field of four quadrants of described micro-visual field microfacies machine adjacent with four respectively shooting gained
Folded.
Further, the size of display visual field must be less than micro-field size of 1/4.
This utilizing camera " ceramic tile " to realize real-time multiple dimensioned formation method, its step is as follows:
Wide visual field camera obtains the thick view field image of whole system visual field;
Microfacies machine obtain system visual field local high-precision micro-view field image, four quadrants of each micro-visual field respectively with
One quadrant overlap of micro-visual field of four adjacent microfacies machine shooting gained.
The display visual field of display can be the most mobile in micro-visual field, when showing a certain particular quadrant, and can
With use cover two of this quadrant micro-magazine any one produce video data.
When showing the next quadrant that visual field moves to current micro-visual field, cover the microfacies machine of current micro-visual field and cover
The microfacies machine of the current new quadrant of lid can provide video data for display visual field jointly.
When user amplifies display visual field, when exceeding micro-visual field of single microfacies machine, wide visual field camera is used to provide display number
According to.
Further, intensive micro-camera array can be used in system field of view center area, it is achieved the quadrant weight of micro-visual field
Folded collection, uses non-overlapped array in the edge zone of system visual field, uses traditional merging algorithm for images.
By utilizing the differences in resolution of microfacies machine and display so that the resolution of microfacies machine is higher than the resolution of display
Rate, and microfacies machine in camera array is staggered so that in micro-visual field of micro-collected by camera gained, each quadrant all has two
Microfacies machine provides video data, during so that display visual field is moved in micro-visual field that the microfacies machine being disposed adjacent is provided,
Obtain real-time high-resolution panoramic image.Additionally, due to the splicing eliminating adjacent cameras image calculates this flow process, not only
Avoid the error that splicing calculates, improve picture quality, also improve display speed, it is ensured that the most clear in system visual field
Look at.
The people being familiar with the industry can be appreciated that this strategy can be applied to multi-level structure from coarse to fine.Will also appreciate that
To the brightness being necessary between matching camera, contrast, gamma curve and other image parameter smoothly to put down from a camera
Move on to another camera.Smooth visual field point by characteristic point between matching camera to move and be also good, this requirement certainly
Looser than the corresponding requirements in traditional images splicing is many.
Accompanying drawing explanation
Fig. 1 is one camera visual field schematic diagram;
Fig. 2 is polyphaser visual field schematic diagram;
Fig. 3 is that visual angle central point arranges schematic diagram;
Fig. 4 is that ceramic tile pyramid shows structural representation;
Fig. 5 is the polyphaser visual field structural representation of the present invention;
Fig. 6 is micro-visual field schematic diagram of the present invention;
Fig. 7 is the system visual field structural representation of the present invention.
Detailed description of the invention
This camera " ceramic tile " is utilized to realize real-time many chis below in conjunction with accompanying drawing and concrete example to of the present invention
Degree imaging system and method are described in further detail.
This camera " ceramic tile " is utilized to realize real-time multiple dimensioned imaging system, including camera array, image procossing and depositing
Send to display after the image procossing that camera array is obtained by storage system, display, described image procossing and storage system;Its
Being characterised by: described camera array is made up of the microfacies machine of some interlaced settings, described microfacies machine shooting the micro-of gained regard
The resolution of field is higher than the resolution of the display visual field of display.
Described camera array also includes a wide visual field camera.
One quadrant weight of micro-visual field of four quadrants of described micro-visual field microfacies machine adjacent with four respectively shooting gained
Folded.
Further, the size of display visual field must be less than micro-field size of 1/4.
This utilizing camera " ceramic tile " to realize real-time multiple dimensioned formation method, its step is as follows:
Wide visual field camera obtains the thick view field image of whole system visual field;
Microfacies machine obtain system visual field local high-precision micro-view field image, four quadrants of each micro-visual field respectively with
One quadrant overlap of micro-visual field of four adjacent microfacies machine shooting gained.
The display visual field of display can be the most mobile in micro-visual field, when showing a certain particular quadrant, and can
With use cover two of this quadrant micro-magazine any one produce video data.
When showing the next quadrant that visual field moves to current micro-visual field, cover microfacies machine and the covering of current micro-visual field
The microfacies machine of current new quadrant can provide video data for display visual field jointly.
When user amplifies display visual field, when exceeding micro-visual field of single microfacies machine, wide visual field camera is used to provide display number
According to.
Further, intensive micro-camera array can be used in system field of view center area, it is achieved the quadrant weight of micro-visual field
Folded collection, uses non-overlapped array in the edge zone of system visual field, uses traditional merging algorithm for images.
We consider a picture system being made up of camera array.We become microfacies the camera in each array
Machine.A part for each micro-camera looks into fee scene.The space of all micro-camera looks into fee is generically and collectively referred to as system visual field, and each microfacies machine is seen
The space examined is referred to as micro-viewing field of camera.We are also contemplated for from camera array the display image produced.Each display image also has
Visual field.Different from spliced panoramic imaging, it will be assumed that each display visual field is corresponding to the subregion of a micro-viewing field of camera.System
Basic structure illustrate in the accompanying drawings.It is presented with micro-camera array and the display image of generation of overlapped fov.
High definition refers in particular to the image size situation significantly more than display device display size in this literary composition.Such as, a 4K figure
As showing on the display of 1920*1080 pixel.Even if in the case of display has higher resolution, by up-sampling
(up-sampling) satisfied display effect can also be presented over the display.Have camera system at present and can gather 100,000,000,
Or even the image of hundred million grades of pixels, these camera systems generally use array camera or by camera is fixed on The Cloud Terrace scanning
Scene, is then spliced into the big figure of panorama.Method described herein uses camera array to promote display effect, but reduce or
Avoid using image mosaic.
The key concept of the present invention is that we require that the resolution of each microfacies machine is higher than needed for showing image
Resolution.We may require that microfacies machine has a more pixel than display, or with interpolation method increase micro-camera image and
Size disparity between display.In both cases, it will be assumed that display image area can be in microfacies machine image range
In up and down, move left and right, in order to make microfacies machine obtain bigger visual field, it is possible to use 4K (3840*2160) camera produces one
Individual standard HD (1920*1080) display image stream.Effect same can by one HD camera of up-sampling generate one equivalent
4K camera image, generates display image stream with this.In both cases, display window can be by selecting in camera window
Respective pixel scope realize moving up and down about numeral.Can be by camera window or down-sampling mates display
Window realizes digital zooming.
The present invention utilizes the method that camera array directly obtains image tiles.With the conventional method utilizing tile structure tree
Different, it would be desirable to tile and display image are fully overlapped can be able to directly obtain aobvious from a camera image stream
Show image stream, it is not necessary to splice.For realizing this target, we are according to organizing camera array visual field shown in accompanying drawing.One wide regards
The thick view field image of the field whole scene of collected by camera.Fine resolution camera array gathers the detail pictures of scene local.Such as accompanying drawing
Shown in, interlaced between fine resolution camera array, four quadrants of each camera and a quadrant weight of four neighbouring cameras
Folded.Require that display window is less than or equal to 1/4 collected by camera window size, it is ensured that display window can be from a camera window
Continuous translation is to next camera window, it is not necessary to splice.When display window is in certain particular quadrant, it is possible to use cover this watt
Magazine arbitrary one of two of sheet produce video data.When display window moves on to next quadrant, cover and work as before-image
The camera of limit and the camera of new quadrant the most smoothly can provide data for display window.In new quadrant, display window can
To transfer to cover equally the new camera of this window, in order to allow display window depart from existing camera, expand to new camera and go.
We can build a combined type framework, uses intensive overlapping collected by camera window in system field of view center area
Mouthful, non-overlapped array is used in system field of view edge area.Translation between edge zone, camera can be by using tradition
Merging algorithm for images and tile mosaic method.As a rule, user is easier to be tolerated in the splicing of edge zone by mistake
Difference.When user amplifies image window, exceedes single micro-viewing field of camera when, use coarse resolution camera.