WO2011069698A1 - Imagerie panoramique - Google Patents

Imagerie panoramique Download PDF

Info

Publication number
WO2011069698A1
WO2011069698A1 PCT/EP2010/064107 EP2010064107W WO2011069698A1 WO 2011069698 A1 WO2011069698 A1 WO 2011069698A1 EP 2010064107 W EP2010064107 W EP 2010064107W WO 2011069698 A1 WO2011069698 A1 WO 2011069698A1
Authority
WO
WIPO (PCT)
Prior art keywords
image
images
panorama
res
relatively low
Prior art date
Application number
PCT/EP2010/064107
Other languages
English (en)
Inventor
Peter Corcoran
Alexandru Drimbarean
Piotr Stec
Larry Murray
Igor Barcovschi
Original Assignee
Tessera Technologies Ireland Limited
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Priority claimed from US12/636,639 external-priority patent/US20110141225A1/en
Priority claimed from US12/636,618 external-priority patent/US8294748B2/en
Priority claimed from US12/636,647 external-priority patent/US20110141226A1/en
Priority claimed from US12/636,629 external-priority patent/US20110141229A1/en
Priority claimed from US12/636,608 external-priority patent/US20110141224A1/en
Application filed by Tessera Technologies Ireland Limited filed Critical Tessera Technologies Ireland Limited
Priority to EP10762898.4A priority Critical patent/EP2545411B1/fr
Publication of WO2011069698A1 publication Critical patent/WO2011069698A1/fr

Links

Classifications

    • G—PHYSICS
    • G03—PHOTOGRAPHY; CINEMATOGRAPHY; ANALOGOUS TECHNIQUES USING WAVES OTHER THAN OPTICAL WAVES; ELECTROGRAPHY; HOLOGRAPHY
    • G03B—APPARATUS OR ARRANGEMENTS FOR TAKING PHOTOGRAPHS OR FOR PROJECTING OR VIEWING THEM; APPARATUS OR ARRANGEMENTS EMPLOYING ANALOGOUS TECHNIQUES USING WAVES OTHER THAN OPTICAL WAVES; ACCESSORIES THEREFOR
    • G03B37/00—Panoramic or wide-screen photography; Photographing extended surfaces, e.g. for surveying; Photographing internal surfaces, e.g. of pipe
    • G03B37/02—Panoramic or wide-screen photography; Photographing extended surfaces, e.g. for surveying; Photographing internal surfaces, e.g. of pipe with scanning movement of lens or cameras
    • G—PHYSICS
    • G03—PHOTOGRAPHY; CINEMATOGRAPHY; ANALOGOUS TECHNIQUES USING WAVES OTHER THAN OPTICAL WAVES; ELECTROGRAPHY; HOLOGRAPHY
    • G03B—APPARATUS OR ARRANGEMENTS FOR TAKING PHOTOGRAPHS OR FOR PROJECTING OR VIEWING THEM; APPARATUS OR ARRANGEMENTS EMPLOYING ANALOGOUS TECHNIQUES USING WAVES OTHER THAN OPTICAL WAVES; ACCESSORIES THEREFOR
    • G03B37/00—Panoramic or wide-screen photography; Photographing extended surfaces, e.g. for surveying; Photographing internal surfaces, e.g. of pipe
    • G03B37/04—Panoramic or wide-screen photography; Photographing extended surfaces, e.g. for surveying; Photographing internal surfaces, e.g. of pipe with cameras or projectors providing touching or overlapping fields of view
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N23/00—Cameras or camera modules comprising electronic image sensors; Control thereof
    • H04N23/60—Control of cameras or camera modules
    • H04N23/68—Control of cameras or camera modules for stable pick-up of the scene, e.g. compensating for camera body vibrations
    • H04N23/681—Motion detection
    • H04N23/6811—Motion detection based on the image signal
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N23/00—Cameras or camera modules comprising electronic image sensors; Control thereof
    • H04N23/60—Control of cameras or camera modules
    • H04N23/698—Control of cameras or camera modules for achieving an enlarged field of view, e.g. panoramic image capture
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N23/00—Cameras or camera modules comprising electronic image sensors; Control thereof
    • H04N23/70—Circuitry for compensating brightness variation in the scene
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N23/00—Cameras or camera modules comprising electronic image sensors; Control thereof
    • H04N23/80—Camera processing pipelines; Components thereof
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N23/00—Cameras or camera modules comprising electronic image sensors; Control thereof
    • H04N23/95—Computational photography systems, e.g. light-field imaging systems
    • H04N23/951—Computational photography systems, e.g. light-field imaging systems by using two or more images to influence resolution, frame rate or aspect ratio

Definitions

  • the invention relates to panoramic imaging with portable and/or handheld cameras, digital still cameras, and other camera-enabled devices such as camera-phones and other handheld devices, and processor-based portable devices with image acquisition capabilities.
  • a panoramic photograph is a photograph with an unusually large field of view, an
  • a horizontal panoramic photograph may be much wider than it is tall, and may have a horizontal angle of view that is large in relation to its vertical angle of view.
  • a vertical panoramic photograph may be much taller than it is wide, and may have a vertical angle of view that is large in relation to its horizontal angle of view.
  • a panoramic photograph or panoramic image can provide a unique and sometimes striking view of a scene.
  • Panorama imaging involves taking a sequence of images of an extended horizontal scene and compositing these into a single extended image. This enables a "panoramic,” typically outdoor, scene to be captured with a standard camera with a normal optical system of limited field-of-view.
  • An alternative approach is to capture a scene with a specialized optical lens known as a fish- eye which has an enhanced field of view of up to 170'.
  • a specialized optical lens known as a fish- eye which has an enhanced field of view of up to 170'.
  • Such specialized lenses require expensive fabrication and precise machining of the elements of the lens.
  • Implementing a panoramic imaging method in a digital camera enables similar results at a fraction of the cost.
  • panoramic photographs have been taken using specially-made cameras.
  • One kind of panoramic camera uses a rotating lens and body to sweep across a large field of view, while moving film past a narrow exposure slit behind the lens.
  • This kind of rotating camera sometimes called a "Cirkut-type" camera after an early commercial model, can take a photograph with a field of view of 360 degrees or more.
  • a swing-lens camera operates on a similar principle, but rotates its lens and the exposure slit in relation to a stationary body and film.
  • a swing-lens camera can take a photograph with a field of view somewhat less than 180 degrees.
  • Another method of making a panoramic photograph may involve taking several overlapping standard or conventional photographs, each typically having an aspect ratio of about 3:2 or otherwise less than is desired in the panoramic photograph, and joining them together into a single larger photograph.
  • the joining may be typically done using a computer operating on digital representations of the component photographs, for example photographs taken with a digital camera.
  • the process of combining digital images into a larger photograph is often called "stitching" or “mosaicing”.
  • any number of component images can be stitched, and the resulting panorama can cover a field of view of up to 360 degrees or more.
  • Stitching can be computationally-intensive.
  • software performing image stitching may correct distortions, such as lens distortion and perspective distortion, that may be present in component images before stitching them together.
  • finding the proper alignment and color or tonal balance between component images may involve multiple computations of correlation coefficients that reflect the "goodness" of the alignment between image segments.
  • Variations in tone, caused by effects such as changes in viewing angle and lens vignetting, may be corrected or otherwise accommodated.
  • the time required to perform the stitching increases dramatically with increased size or resolution of the component images.
  • a panoramic mode may use a display screen on the camera to assist the user in framing each component photograph for proper overlap with a previous photograph in the set, and may ensure consistent exposure settings for all of the component photographs in a set.
  • At least one existing model of digital camera can perform stitching on a set of low-resolution "screen nail” images so that the photographer can detect certain problems such as insufficient overlap in the component images.
  • a "screen nail” may include a small low-resolution copy of a digital image, analogous to a "thumbnail” image, and is sized to fit an on-camera display.
  • Atypical screen nail image may have, for example, approximately 320 by 240 pixels. This capability is described in U.S. patent application 20060182437 (Hewlett-Packard).
  • US 20090022422 to Sorek et al. describes a method for combining at least a first and second image frame based on the content of the frames and combining these frames in a transformed domain based on a determined alignment to form a composite image.
  • This method employs common features of interest in each image and uses these to determine the horizontal and vertical alignment between the two images.
  • the combining of the images may be partly performed in the image domain where residual shifts are performed and partly in the transform domain where block shifts may be performed.
  • an exposure level is fixed for acquiring the panorama image with the portable imaging device.
  • the imaging device is panned across a panoramic scene. At least two relatively low-resolution image frames of overlapping portions of the panoramic scene are acquired and processed.
  • the processing includes sorting and retaining a set of relatively low resolution image frames. A relative displacement between each image of the set of relatively low resolution image frames is determined. An approximately optimal stitch line is also determined between each pair of images of the relatively low resolution image frames.
  • the method also includes acquiring and storing a set of main image frames corresponding to the set of relatively low resolution image frames.
  • the main image frames are registered or aligning, or both, based on relative displacements of corresponding images of the set of relatively low resolution images.
  • One or more approximately optimal stitch lines determined for lo-res image pairs is/are mapped onto one or more corresponding pairs of registered and/or aligned main image frames that are joined to form a main panorama image.
  • the determining of an approximately optimal stitch line may include determining an alpha blending map in the vicinity of the approximately optimal stitch line.
  • the mapping of the approximately optimal stitch line onto the high resolution images may include further mapping the blending map, wherein the joining of the high resolution images includes blending the images based on the mapping of the alpha blending map.
  • the joining may include blending the set of main image frames, including mapping the alpha-blending map for a series of the relatively low resolution image frames to a main series of image frames.
  • One or more of the component image frames that are joined to form the panorama image may be cropped.
  • a set of two, three or more low resolution image frames may be acquired of component portions of the scene.
  • a user may be notified and/or an image discarded when a horizontal overlap between consecutive images falls outside a predetermined range.
  • the method may also include performing sub-pixel image registration on the relatively low resolution images to prevent pixel offsets in the mapping of the alpha-blending map for the relatively low resolution series to the main series.
  • the joining of pairs of digital images of the main series may be interleaved with the sub-pixel registration of relatively low resolution images and/or with the acquiring and/or generating, and/or joining of corresponding pairs of relatively low resolution images.
  • a further method of generating a panorama image using a portable imaging device includes fixing an exposure level for acquiring the panorama image using the portable imaging device, and panning the imaging device across a panoramic scene.
  • a set of at least two image frames of overlapping portions of said panorama image are acquired and processed using an optic and imaging sensor of the portable imaging device.
  • the processing includes sorting and retaining a set of image frames including one or more overlapping pairs of image frames.
  • a relative displacement is determined between each of the set of overlapping image frames, including determining an overlapped region for each image pair.
  • the images of the set are registered and/or aligned based on the relative displacements.
  • An alpha blending map and/or an optimal stitch line is/are determined for each pair of overlapping image frames.
  • the one or more pairs of image frames are joined to form a panorama image that is stored, transmitted and/or displaying.
  • the determining of relative displacement may include determining relative horizontal displacement between a pair of images of a set of overlapping image frames. Pixel values in image columns may be summed across each of first and second images to determine a horizontal image profile for each image. A column difference profile may be determined across each of images. A relative error function may be determined between the pair of images according to image column difference profiles. A minimum of the relative error function indicates a relative number of pixel columns of horizontal displacement between the pair of images.
  • the determining of relative displacement may further include determining relative vertical displacement between the pair of images of the set of overlapping image frames. Pixel values in image rows may be summed across each of the first and second images to determine a vertical image profile for each image. A row difference profile may be determined across each of the images. A relative error function may be determined between the pair of images according to image row difference profiles. A minimum of the relative error function indicates a relative number of pixel rows of vertical displacement are between the pair of images.
  • a smoothing function may be applied to the column and/or row difference profiles of each image prior to calculating the relative error function between the pair of images.
  • the joining is performed using the alpha blending map or the optimal stitch line, or both.
  • the joining may be based on determining both the approximately optimal stitch line and the alpha blending map.
  • the approximately optimal stitch line may include an approximately 50% blending ratio between overlapped pixels of the pair of images, and the map may provide a blending ratio for overlapped pixels from the pair of images in the vicinity of the approximately optimal stitch line.
  • the method may also include interleaving joining of pairs of image frames with acquiring and/or generating of next image frames.
  • the panorama image may also be cropped to a uniform vertical height.
  • a super-resolution method for generating a panorama image using a portable imaging device including an optic, image sensor and processor.
  • An exposure level is fixed for acquiring the panorama image with the portable imaging device.
  • the imaging device is panned across a panoramic scene.
  • Two or more sets of images are acquired are processed, each including at least two image frames of portions of the panoramic scene.
  • the method includes sorting and retaining multiple images of the two or more sets.
  • a relative displacement is determined between each pair of neighboring frames within each of the two or more sets of image frames.
  • the method also includes registering images within each of the image sets relative to one another.
  • Each of said two or more sets is joined to form two or more substantially overlapping panorama images. These are combined to form a higher resolution panorama image of substantially the same scene, which is storied, transmitted and/or displayed.
  • the combining of the two or more substantially overlapping panorama images may include applying a super-resolution technique.
  • the cropping of the higher resolution panorama image may include removing one or more non-overlapping regions from one or more component panorama images.
  • the method may also include determining a relative displacement between a first or otherwise corresponding acquired frame of each of the two or more sets of image frames.
  • a combined panoramic image derived from images of each of the two or more sets may be registered.
  • the method may include interleaving joining of pairs of image frames with acquiring and/or generating of next image frames.
  • Another method of generating a panorama image including panning across a scene a processor-based device configured for acquiring digital images.
  • multiple main series images are acquired with the device.
  • Each of the multiple images contains a different angular range of the panorama scene.
  • relatively low resolution images are acquired corresponding to the multiple main series images and/or the main series images is subsampled to produce the relatively low resolution images.
  • the relatively low resolution images are joined to form a low-res panorama, which is displayed.
  • the multiple main series images are composited in real-time on the device based on the joining of the relatively low resolution images to form a main series panorama image.
  • the joining may include stitching and/or matching exposure, color balance, or luminance, or combinations thereof, aligning and/or registering edge regions of images, and blending matched, aligned images.
  • the blending may include blurring a seam line generated between adjacent component images of the panorama image.
  • An exposure level may be fixed for acquiring the panorama image with the portable imaging device.
  • the joining may include aligning and/or registering edge regions of images, and blending matched, aligned images.
  • the blending may include blurring a seam line generated between adjacent component images of the panorama image.
  • the compositing may include estimating a global motion of the panning and determining whether the global motion is sufficient.
  • the compositing may include estimating a relative scene displacement and determining whether the relative scene displacement is sufficient.
  • the method may include notifying a user, discarding one or more of the images and/or interrupting the method, when the relative scene displacement is determined to be insufficient.
  • the relative scene displacement may be determined to be sufficient when frame to frame overlap comprises a range between 10% and 40%), or between 20%> and 30%>, or insufficient if outside a predetermined range such as 10- 40%, 20-30%, or otherwise.
  • the estimating relative displacement may include multidimensional motion estimating for multiple image pairs.
  • the method may be performed with or without user intervention, and with or without device motion measurement.
  • the panorama image may be compressed prior to storing.
  • the joining of the relatively low resolution images may include generating an alpha blending map, and the alpha blending map may be used in the joining of the main series images.
  • a further method for generating a panorama image.
  • the method involves using a processor-based image acquisition device configured both for acquiring a main series of digital images of a panoramic scene and generating and/or acquiring a series of relatively low resolution images corresponding to the main series.
  • the device is panned across a panoramic scene.
  • the main series of digital images is acquired with the device.
  • Each of the main series of digital images contains a different angular range of the panoramic scene.
  • the series of relatively low resolution images corresponding substantially to the same panoramic scene as the digital images of the main series are acquired and/or generated.
  • Images of the series of relatively low resolution images are joined to form a relatively low resolution panoramic image.
  • a map of the joining of the series of relatively low resolution images is generated.
  • the main series of digital images is joined based on the map to form a main panoramic image, which is displayed, stored, further processed and/or transmitted.
  • the method may further include estimating a relative scene displacement during the panning. Portions of the panoramic scene may be selectively captured during the panning based at least in part on the estimating of the relative scene displacement.
  • the method may include notifying the user, discarding one or more of the images and/or interrupting the method, when the relative scene displacement is determined to be insufficient, e.g., when frame to frame overlap is determined to be outside of a range between 10% and 40%, or 20%> and 30%>, or other set or selected range.
  • the estimating relative scene displacement may involve multi-dimensional displacement estimating for multiple image pairs. Horizontal (long panorama image dimension) and vertical offsets may be determined between consecutive images in the series based on the estimating of the relative scene displacement.
  • An image of the series may be discarded if it has less than a threshold horizontal offset from a previous image of the series.
  • the user may be notified when the panning of the device does not exceed a threshold motion.
  • the user may also be notified, and/or an image of the series may be discarded, when a vertical offset exceeds a threshold offset with another image of the series.
  • the joining of main series and relatively low resolution images may include stitching.
  • the stitching may involve aligning and/or registering edge regions of images, and blending matched, aligned images.
  • An exposure level may be first for acquiring the panorama image with the portable imaging device, and/or the joining may involve matching exposure, color balance, and/or luminance.
  • the blending may include blurring a seam line generated between adjacent component images of the panorama image.
  • the map may comprise an alpha-blending map including information of approximately optimal stitching seams determined from the joining of the relatively low resolution images.
  • the joining of the main series may include mapping of the alpha-blending map for the low resolution series to the main series.
  • the method may involve sub-pixel image registration of the relatively low resolution images to prevent pixel offsets from occurring during the mapping of the alpha-blending map for the low resolution series to the main series.
  • the joining of the series of relatively low resolution images and/or of the main series may be based in part on the estimating of the relative scene displacement.
  • a portable camera-enabled device capable of in-camera generation of a panorama image including a lens, an image sensor, a processor, and a processor readable medium having code embedded therein for programming the processor to perform any of the panorama image generation methods described herein.
  • Processor-readable storage media are also provided that have code embedded therein for programming a processor to perform any of the panorama image generation methods described herein.
  • a method is presented of generating a high resolution panorama image within a digital imaging device which can capture either HD video or high resolution still images.
  • the method is performed without user intervention and without a requirement for external measurement of motion of the imaging device. Further the method is performed in "real- time" and with consideration for the limited memory resources on such devices which can generally store small numbers of high resolution video frames or still images at a time.
  • the method involves the user panning the device at a natural speed across a scene of which it is desired to capture a panoramic image.
  • the device displays the scene as the user pans across it, in the same way that it would when capturing a normal video sequence.
  • the method involves the capture or generation of a lower resolution sequence of images than the main high-resolution video or still images.
  • a camera-enabled device may incorporate hardware which generates a "preview stream", a sequence of low resolution images which is typically used to provide a real-time display of the data captured by the imaging sensor.
  • a device may still have an "image subsampling unit" which can generate, almost instantly, a lower resolution version of the full image or video frame.
  • Figure 1 illustrates features of a digital imaging device configured to generate full resolution main images and subsampled or resized images from a main image acquisition and processing chain in accordance with certain embodiments.
  • Figure 2 shows example plots of horizontal image profiles for two image frames, including plotting the sums of columns for each row for each of the two image frames, wherein the plots shown are substantially overlapped.
  • Figure 3 shows example plots of the profiles of Figure 2 after calculating differences along the profiles, wherein again the plots shown are substantially overlapped.
  • Figure 4 shows a plot which illustrates motion estimation between the two image frames of Figures 2 and 3.
  • Figure 5 shows a plot of pixel shift versus frame number for horizontal camera movement measured from video or sequential image frames.
  • Figure 6 shows a plot of pixel shift versus frame number for vertical camera movement measured from video or sequential image frames.
  • Figure 7A illustrates three images stitched together from sequential images taken with a camera moving substantially horizontally and slightly vertically between frames, wherein each of the three images overlaps one or two adjacent images slightly horizontally and substantially vertically.
  • Figure 7B illustrates a panorama image generated by cropping in the vertical direction two or more of the three images stitched together of Figure 7A.
  • Figure 8 is a plot that illustrates comparing a reference profile with interpolated values of a shifted profile in accordance with certain embodiments.
  • Figure 9 is a plot illustrating changes of error between profiles in respect to sub-pixel shift in accordance with certain embodiments.
  • Figure 10 illustrates an evolution of two contours propagating towards each other across an overlap area as part of a process in accordance with certain embodiments.
  • Figure 11 shows a panoramic image generated from two images and blended by a process in accordance with certain embodiments.
  • Figure 12 illustrates a blend mask used in the generation of the panoramic image shown at Figure 11.
  • Figure 13 illustrates an image blending process in accordance with certain embodiments.
  • Figure 1 illustrates a digital imaging device configured to capture full resolution main images and to generate subsampled (resized) images from the main image acquisition and processing chain.
  • a sensor 102 captures a full-res image, and the image data is applied to sensor processing 104 and an imaging pipeline 106.
  • An image subsampler 108 generates
  • the hires images may be JPEG compressed at 110.
  • the compressed, full-size images 111 may be stored in a memory 112, which may be a temporary image store 112, along with the subsampled images 114.
  • a lo-res panorama image 116 may be generated, and stored temporarily in the image store 112, by joining two or more of the subsampled images 114.
  • the low-res panorama image 116 may or may not be displayed.
  • An image post-processor 118 generates a hi-res panorama image 120 based on information gathered in the generation of the lo-res panorama image 116, thereby advantageously saving computing resources.
  • the hi-res panorama image 120 may be stored in an image store 122, such as a SD card or equivalent, along with the full-size images 111.
  • the low-resolution images may be processed in accordance with another embodiment as follows. Firstly, an initial lo-res image or video frame may be obtained and used as a first reference frame. The corresponding hi-res image is stored. Next, additional lo- res images may in certain embodiments be obtained, and an optional method of global motion estimation, e.g., such as that described at US published patent application 2008/0309769, may be applied between the reference frame and each additional image frame to determine their horizontal and vertical offsets. If the horizontal offset is less than a predetermined range, then the lo-res frame may be discarded along with the corresponding hi-res image. In certain embodiments (e.g. still camera), the acquisition of the corresponding hi-res image may not yet be completed. In such cases, hi-res image capture may simply be aborted and a new acquisition may be initiated.
  • an optional method of global motion estimation e.g., such as that described at US published patent application 2008/0309769
  • the process may be halted and/or an error message may be displayed, e.g., such as "camera not panned by user.”
  • a warning beep may indicate to the user to pan faster.
  • an error indication may be provided to warn the user that they are "drifting.” In certain embodiments, no user direction is involved, while the user pans the device across a scene at a reasonable speed in the panorama process.
  • the frame-to-frame overlap may be set, for example between 10-30% and 20-40%, or between 20%> and 30%> in one embodiment, then that frame is retained and the corresponding hi-res image is stored. This retained frame becomes the next reference image and the process is repeated.
  • the acquisition process is halted.
  • the low-res images are next "joined” using a method which is particularly effective for low-resource embedded devices.
  • the resulting low resolution image is then cropped and may optionally be displayed for user acceptance.
  • an alpha- blending map may be created for the joining of the overlapping regions between adjacent pairs of low-res or reference image frames in the panorama. This same map may then be advantageously used to join corresponding hi-res images to create a hi-res panorama image.
  • This alpha-blending map means that joining algorithms do not have to be repeated for the hi-res images, which would otherwise be resource intensive, because an "optimal seam" between each image pair has advantageously already been determined from the joining of the lo-res images.
  • a method of performing sub-pixel image registration on lo-res images is also provided in certain embodiments to ensure that pixel offsets do not occur when mapping the alpha- blending map from low-res to hi-res image pairs.
  • the hi-res panorama image is then compressed to JPEG and stored.
  • the joining of low-res pairs may be interleaved with the acquisition of new lo-res image frames.
  • the joining and JPEG compression of portions of the hi-res images may also be interleaved with the acquisition, registration and joining of lo-res images.
  • Image stitching as applied to the generation of a panorama image in accordance with certain embodiments, may involve any or all of the following steps:
  • Image calibration including perspective correction, vignetting correction, and/or chromatic aberration correction, wherein images may be processed in this optional stage to improve results.
  • Image registration including analysis for translation, rotation, and/or focal length, wherein direct or feature-based image alignment methods may be used.
  • Direct alignment methods may search for image orientations that minimize the sum of absolute differences between overlapping pixels.
  • Feature-based methods determine proper image orientations by identifying features that appear in multiple images and overlapping them.
  • Image blending, or otherwise combining the sections may involve any or all of: color correction, including matching adjoining areas of component images for color, contrast and/or brightness to avoid visibility of the seams; dynamic range extension; and/or motion compensation, deghosting, and/or deblurring to compensate for moving objects.
  • image registration may involve, in certain embodiments, determining relative displacement of images based on global image analysis of image rows and columns, rather than on localized pixel-by-pixel analysis which may involve a calculation to be taken for pixels of each adjacent image.
  • step 3 color correction and/or analysis of local contrast and/or brightness levels may be eliminated in embodiments involving determining an alpha blending map and/or an image seam in a single operation.
  • techniques in accordance with certain embodiments may be performed without certain otherwise standard steps used in conventional image stitching algorithms, and/or while simplifying certain steps to provide a method particularly suited to
  • stitching technique in accordance with certain embodiments may be referred to as “joining,” because it differs so greatly from conventional stitching algorithms.
  • an exposure level is fixed on the imaging device prior to acquisition of the main set of images to be processed.
  • the level may simply be fixed at a level suitable for acquisition of the initial frame of the panorama sequence.
  • the user may be allowed to manually increase or reduce the exposure level.
  • the user may be prompted to perform a prescan of the panorama scene so that an average exposure level across the entire scene may be determined. This "prescan" may involve sweeping the camera-phone across the panorama scene prior to the main acquisition sweep.
  • a second phase of panorama creation in accordance with certain embodiments involves the user sweeping or panning the device at a natural speed across a scene of which it is desired to capture a panoramic image.
  • the device may optionally display the scene in real-time as the user pans across it, in the same way that it would when capturing a normal video sequence or when composing a still image.
  • the acquisition process may terminate in any of a number of ways: by detecting that there is (i) no change (or change below a threshold) between following image frames (user holds camera in a fixed position); (ii) no similarity (or similarity below a threshold) between following image frames (user very rapidly moves camera to a different field of view); (iii) a sudden motion of the camera (for cameras equipped with motion sensor); (iv) a switch depressed or a "dead-man" switch released; (v) elapse of a time interval; (vi) a threshold number of main image frames acquired or saturation of the memory capacity of the device or various combinations of the above events.
  • the camera-phone While the camera-phone is acquiring images during the panorama sweep, it may be constantly processing these images so that a full panorama image may be acquired in real time. During this processing, many of the acquired images may be discarded in a sorting process in accordance with certain embodiments, while those which are relevant to the final panorama are retained. Furthermore these images may be registered in real time. A blending operation may be performed on the registered images.
  • the sorting, registration and initial blending may be performed on subsampled versions of the main acquired images or video frames.
  • the sorting process may enable corresponding main image frames to be retained or immediately discarded so that a small number of full sized images are stored during the panorama sweep.
  • the registration and blending process may enable an optimal "seam line" between lo-res images to be determined as images are acquired. As soon as the panoramic sweep is terminated, retained full-res images may be joined together without any delay using the registration and optimal seam information from the corresponding lo-res or preview images.
  • the resulting hi-res panorama image may then be compressed to JPEG format and stored.
  • the joining of low-res pairs is interleaved with the acquisition of new lo-res image frames.
  • one or more segments of the joining and/or JPEG compression of portions of the hi-res images may also be interleaved with the acquisition, sorting, registration and/or blending of lo-res images.
  • x & y pixel displacements are determined of one image frame relative to the next. This displacement information can then be used to select the image frames that will be used in the creation of a panorama image.
  • US20060171464 and US20080309769 are referenced here.
  • Methods of video stabilization are described using image-processing techniques. Techniques in accordance with certain embodiments involve estimating inter-frame horizontal and vertical displacements of one or more pairs (or a sequence) of image frames.
  • This image frame displacement information can be used to select an image frame which will overlap the previous image frame by a desired number of pixels. In certain embodiments, it may alternatively be used to direct the camera to acquire (or complete acquisition of) a new full-resolution image.
  • Profiles of images 1&2 are summed and then smoothed using a running average kernel of length 15 in an exemplary embodiment.
  • the length may be variable, e.g., such that the length can be increased to further smooth profiles.
  • the image pixels values are summed along columns and the sums from the columns form a vector that may be referred to herein as a profile in accordance with this exemplary embodiment.
  • a profile For the sake of simplicity, only G channel is used in case of RGB image and Y channel in case of YUV image. Any colour channel containing enough information about image details can be used in that case.
  • the profiles from both registered images are smoothed and differentiated using a convolution kernel of length adapted to the image size and amount of noise. Length is variable is certain embodiments, and can be increased or decreased.
  • Horizontal (X) displacement estimation sums all the columns and vertical (Y) displacement estimation sums along the rows of the image. Thereafter the process of motion estimation is the same therefore only details for horizontal displacement are outlined below. Initially all of the columns of each row are summed for image 1 and 2. This is to create image profiles from the displacement estimation.
  • Figure 2 illustrates image profiles when columns are summed and after differences are calculated.
  • the next step is to differentiate and smooth the profiles. This has the effect of making the displacement estimation invariant to intensity differences between the images.
  • horl diff (horl) ;
  • hor2 diff (hor2) ;
  • m is the number of rows.
  • Figure 3 illustrates image profiles after differences are calculated.
  • Figure 3 shows plots of horizontal image profiles as summed values of columns versus row numbers. The profiles are then smoothed to reduce the effects of noise with the following MATLAB code.
  • MATLAB a convolution function is used which is equivalent to a running average filter is shown below,
  • avg ones ( 1 , kernel ). /kernel ; %Running average kernel
  • horl conv (horl , avg) ;
  • horl horl ( kernel/2 : end-kernel/2 ) ;
  • hor2 conv (hor2, avg) ;
  • hor2 hor2 (kernel/2 : end-kernel/2) , ⁇
  • herr zeros ( 1 , 2*radius+l ) /
  • one image profile is shifted relative to the other in 1 pixel shifts.
  • the sum of the absolute differences is calculated to find the shift that minimises the differences.
  • herri i-30i+30horl-hor2 (i) wl
  • Figure 4 illustrates motion estimation by plotting the sum of absolute differences found by shifting profiles versus pixel shift.
  • Figure 4 illustrates image profiles after differences are calculated. The position of the minimum in Figure 4 indicates displacements between profiles.
  • both the x and y motion estimates are calculated for the global value and for the 4 sub-regions described above.
  • the method used to filter out errors due to subject motion within the scene is described below. In this part there are no calculations to describe just logic operations.
  • Figure 5 illustrates the cumulative x-shift in pixels along the horizontal direction with the image frame number.
  • Figure 5 illustrates a cumulative plot showing the total horizontal movement of the camera as pixel shift versus frame number.
  • Figure 6 illustrates a cumulative plot showing the total vertical movement of the camera as pixel shift versus frame number.
  • this displacement measuring technique to also provide a sorting mechanism for which images should be retained we retain displacement information from all earlier image frames and use these to adjust the search area for subsequent images.
  • information from frames 1 and 2 is used to adjust the search area for images 1 and 3. This is done until the total x-shift from image 1 to image 1+n exceeds a given number of pixels. At this point image 1+n becomes a new reference frame and subsequent shifts are registered relative to it. All intermediate frames 2, 3 ... n may be dropped and only frames 1 and 1+n are retained. This process is repeated until some event terminates the panorama acquisition process.
  • the displacement value for the sorting process is chosen to ensure that there is a sufficient overlap between images that are to be joined together. It may be varied also to account for the lens or optical distortions on image frames. Thus where there is significant edge distortion a greater overlap may be required between image frames.
  • Figure 7A illustrates a panorama image generated from video frames chosen by algorithm, and aligned with the calculated x & j-shifts in accordance with certain embodiments.
  • Figure 7B illustrates a cropped version of the panoramic image of Figure 7A.
  • An image registration method in accordance with certain embodiments can compute horizontal and vertical displacements with at least an accuracy of one pixel.
  • the algorithm is configured to prevent what would be otherwise noticeable errors across portions of the seam line apparent when conventional algorithms are used.
  • a one pixel registration error at Q VGA resolution can translate to 6 pixels error for 720p HD resolution.
  • an image registration method in accordance with certain embodiments was extended to compute image displacements with sub-pixel accuracy, i.e., with fractional precision.
  • Estimation based on spline oversampling of profiles may be used for aligning images with high pixel accuracy.
  • profiles may be initially aligned to one pixel accuracy.
  • An oversampling step may be used to find displacement in a range ⁇ 1 pixel with a predefined sub-pixel step.
  • a current implementation may assume a minimum step of l/256th of a pixel. Thanks to this spline, coefficients can be pre-calculated and stored in a lookup table with 256 elements.
  • the minimum step value may be used as a unit during sub-pixel alignment.
  • the step size for estimation can be set to 64.
  • the second profile is shifted left and right with step increments and the values in-between the samples are calculated using spline interpolation. Since the shift is uniform for all the profile samples, it may involve as little as a single set of coefficients to process all the samples.
  • One of the steps is illustrated at Figure 8.
  • Figure 1 shows a plot comparing profile 1 to interpolated values of profile 2 in an exemplary embodiment. Interpolated values are marked with squares and the reference values marked with circles are used to calculate sum of errors for given shift. The shift that produces the least sum of errors is returned in fixed point format with 8 bits fractional value. The next picture shows an example of error versus shift plot.
  • Figure 2 illustrates error metric values versus sub-pixel shifts in accordance with certain embodiments.
  • the error function is smooth within the considered range and without local minima which guarantees finding the correct fractional shift.
  • HD frames are down-sampled to QVGA
  • First QVGA frame is considered reference frame
  • QVGA frames are registered with a reference frame in accordance with certain embodiments.
  • an alpha-blending map may be created by applying a blurring kernel along the seam line;
  • the alpha-blending map may be up-sampled to HD frame resolution
  • HD frames are blended together using the map generated at step 5;
  • the general idea is to have two contours initiated at opposite sides of the overlap area that will move towards each other with a speed which is dependent on the local difference between the overlapping images.
  • the contours propagate across the image horizontally with a certain elasticity so that certain pixels may move more quickly, although they are held back if neighbouring pixels have a lower speed.
  • the idea is that pixels move quickly through regions where there are large differences between the overlapping images and slow down where there is a small or negligible difference between the images. In this way, the two contours which propagate horizontally towards each other meet at an approximately optimal or near-optimal seam between the two images.
  • pixels Once pixels have been visited by the contour, they are assigned to the relevant image associated with the contour. Thus, the contour propagating from left-to-right will associate pixels with the left-hand image. The contour moving from right-to-left will associate pixels with the right-hand image of the overlapping image pair. The process stops where there are no more unassigned pixels in the overlap area.
  • Two overlapping component images are deemed to be overlapping in the horizontal direction or in the long direction of the panorama or direction of panning. So, there is a region which only belongs to the left hand (LH) image, and a region which only belongs to the right hand (RH) image. Then there is an overlapping region where pixels can be chosen from either the LH image or the RH image or some combination or calculation based thereon. Each contour starts at the LH or RH boundary of this overlapping region.
  • the LH boundary belongs entirely to the LH image, while the technique involves moving this boundary to the right along the pixel rows of the image and at the same time moving the RH boundary to the left.
  • the (relative) speed with which each boundary moves is determined by the (relative) difference between the overlapping LH and RM image pixels. Where there is very little difference between a RH and LH pixel, the contour moves more slowly because this is a better place to have an image seam than a faster moving segment. Where there is a large difference, it moves quickly because a seam should not occur in a region of high differences between RH and LH image pixels. In the overlap region, a pixel can be chosen from either the LH or RH image in certain embodiments. Also note that "speed" here is relative among the segments of the contours within the overlapping region.
  • the contour pixels in certain rows will be propagating faster than in the row above (or below), because of variations in the pixel differences.
  • the speed at which a row propagates to the right (for LH contour) or to the left (for RH contour) is also partly constrained by its neighboring rows in certain embodiments.
  • the contour is referred to as being "elastic".
  • one row would be moved very quickly based on large differences between RH and LH images, it may be slowed down because the rows above (or below) are moving at a significantly slower rate, and vice-versa.
  • Alternative embodiments may use techniques involving propagating contours in one direction (contour cannot go back) similar to those described by Sethian et.al, A Fast Marching Level Set Method for Monotonically Advancing Fronts, Proceedings from the National Academy of Sciences, submitted October 20, 1995.
  • an approach similar to the Dijkstra algorithm may be utilized, for example, as set forth in Dijkstra, E. W. (1959). "Anote on two problems in connexion with graphs”. Numerische Mathematik 1 : 269-271, http://www- m3.ma.U m.de/twiki/pub/MN0506/WebHome/Uijkstra.pdf, and/or Sniedovich, M. (2006). "Dijkstra's algorithm revisited: the dynamic programming connexion” (PDF). Journal of Control and Cybernetics 35 (3): 599-620,
  • Contour evolution may be governed by sorting values obtained by solving differential equations. It is relatively fast algorithm on a PC, but involves repetitive sorting (at each contour update by 1 pixel), and dynamic data structures. As such, certain embodiments particularly involving embedded platforms do not use the Dijkstra approach.
  • a FIFO queue is used to implement and realize an advantageous technique, even without memory reallocation.
  • Figure 10 illustrates two contours propagating towards each other across an overlap area between adjacent images to be used in a panorama.
  • the propagation of the contours may be controlled by a queue containing a set of 256 prioritized FIFO buckets.
  • Bucket number 255 has a highest priority while bucket number 0 has a lowest priority.
  • Bucket number 255 has a highest priority while bucket number 0 has a lowest priority.
  • Each point added to the queue is tagged with a flag that identifies the contour the point belongs to.
  • An exemplary propagation algorithm may proceed as follows:
  • the shape of the contour can become very intricate which means the contour has a greater length that requires a larger amount of memory to maintain the queue.
  • the difference between overlapping images may be smoothed by a filter defined by the following matrix:
  • This filtering operation may be equivalent to a blurring of the seam line with a 3x3 blur kernel.
  • other blurring techniques may be used or the size of the kernel may be increased to 5x5 or other values.
  • An example of a full panorama created from two pictures is provided at Figure 11.
  • Figure 11 illustrates a panorama created from two images and blended in accordance with certain embodiments.
  • the seam line can be calculated in reduced resolution using
  • the overlapping parts may be padded with extra pixels, for example to allow integer reduction factors. Padded parts are not taken into account in certain embodiments, and instead are each directly assigned to a proper image. In certain embodiments where the joined image is not to be upscaled, the blurring of the seam line may be omitted.
  • the join map size may be in certain embodiments exactly the reduced size of the image overlap area extended by padding. After calculation of the join line, the map may be enlarged by the same factor as was used for reduction.
  • a greyscale alpha blending map is created by blurring the seam line using a 3x3 Gaussian kernel. Both seam line and alpha blending maps are then up-scaled to a certain output resolution. In one embodiment, up- scaling is achieved using bilinear interpolation. This embodiment uses the alpha blending map to join the high resolution images, thereby avoiding a "boxy" seam. Thanks to the blurry transition of the blending map, the join line becomes less visible at full resolution. In certain embodiments where the joined image is not to be upscaled the blurring of the seam line may be omitted.
  • Figure 12 illustrates a blend mask used to create the panoramic image illustrated in Figure 11.
  • the seam created with this method is invisible or hard to notice as long as the brightness of the two joined images is consistent. Variations in brightness are avoided by fixing the exposure of the imaging device prior to capturing the image set which is used to composite the final panorama image. In certain embodiments, techniques which employ matching sharp and blurred image pairs may be employed to overcome variations in brightness across the image set. Such
  • embodiments may employ techniques drawn from US2008/0219581 or USSN 12/485,316.
  • SET short exposure time
  • preview images within a single blurred main image acquisition may be used to create multiple video frames.
  • the camera-user may be prompted to perform a pre-acquisition sweep of the panorama scene so that an optimal exposure level may be determined across the entire scene. The user then returns the camera to the original location and the exposure level is fixed based upon the pre-acquisition sweep of the scene.
  • the exposure may be varied as each image is acquired and a detailed "delta-map" recording the relative exposure differences between images is recorded and employed to perform tone-balance and/or color matching prior to determining the optimal seam line.
  • color matching may be integrated within the imaging chain and may take adjustment data directly from the exposure "delta-map".
  • the placement of the seam between two images of the panorama can be easily controlled by modifying image-differences calculated between pixels of the overlap region between these two images.
  • the output of a face tracking algorithm can be used to select regions where an image seam is not desirable.
  • any "face regions" within this overlap region may have the image difference artificially increased to decrease the likelihood of, or completely prevent, a contour propagating algorithm from determining a seam which cuts across a face region.
  • a similar logic may apply to foreground portions of an image scene where image details are more perceptible to a viewer. It is less desirable in certain embodiments for an image seam to cross such a region.
  • outputs from a real-time face tracker or a foreground/background detector are available to the panorama algorithm and the image joining operation is modified to accommodate the outputs of the face tracker and/or foreground/background region detector.
  • foreground regions, objects and/or detected face regions which lie in the overlap region between two images may be marked so that a seam line does not cross such regions.
  • the seam is not completely prevented from crossing the determined region, but image differences are artificially increased to make this less likely.
  • the artificial increase in image difference is based on a depth map with a seam being progressively more likely to cross objects that are closer to a distant background of the image.
  • a face or foreground region, or object extends across an entire overlap region between two images
  • an error condition may be signalled and the acquisition process may be aborted.
  • a more sophisticated algorithm may be applied which incorporates face beautification techniques (see, e.g., US patent applications serial numbers 12/512,843, 12/512,819, and 12/512,716).
  • face beautification techniques see, e.g., US patent applications serial numbers 12/512,843, 12/512,819, and 12/512,716).
  • Such algorithm may include texture analysis of the skin followed by subsequent feature- aware smoothing of skin regions.
  • knowledge of principle facial features e.g., eyes, nose, mouth and hairline, combined with knowledge of an external face contour, e.g., chin and sides, may be employed.
  • these may be further refined by knowledge of in-plane face orientation and face pose.
  • knowledge may be obtained using Active Appearance Models or rotational classifiers as described in US patents 7,565,030 and 7,440,593, and US published application numbers US20090179998, US20070269108, US2008/0219517, and US2008/0292193, and US patent application serial no. 61/221,425.
  • parallax effects may occur within the overlap region.
  • an error condition may be signalled.
  • the foreground object(s) and background scene may be joined separately and the foreground re-composited over the background scene.
  • information from additional images overlapping the same overlap region may be employed to facilitate an accurate joining of the foreground object and to determine an optimal positioning of the object within the joined background scene.
  • Information from additional images may also be used to extend the imaged background, and provide details that were hidden by the foreground object due to parallax.
  • one or more foreground object(s) include one or more human faces, and possibly portions of associated body profiles (e.g., head & shoulders, or perhaps more of the torso and arms, and even legs).
  • at least one of the faces is determined to be disposed within the panorama in an overlap region between the two image frames. Due to parallax, this foreground silhouette will appear to be displaced between the two images. If normal joining and/or stitching is applied, the silhouette would be elongated in the horizontal direction producing a "fat face" effect. In a more extreme example where the person is even closer to the imaging device, a "mirror image" of the person may be created in the final joined panorama image. To avoid these undesirable effects,
  • foreground/background separation is applied to each of the component images, e.g., in accordance with any of US patents 7,469,071 and 7,606,417 and US published applications 2009/0273685, 2006/0285754, 2007/0147820, 2009/0040342 and 2007/0269108.
  • the foreground region(s) may be cut from each of any two adjacent images to be joined and/or stitched. In one embodiment, these "cut-out" regions are filled-in using data from other image frames which overlap the same silhouette region, yet expose a different portion of the background due to a different parallax. Additional infilling and extrapolation techniques may be employed where sufficient data is not available.
  • the background images may then be joined using the normal algorithms.
  • the foreground regions may also be separately aligned (e.g., for face regions by matching the locations, and optionally shapes, of facial features such as the eyes, mouth, hairline and/or chin. Following alignment, these may then be joined using the usual algorithms.
  • one of the foreground regions may be selected for compositing onto an already joined background image.
  • the location at which the foreground object is composited onto the background may be selected at a point intermediate or otherwise between the two original locations. The user may even select to position the foreground image even in a different location within the panorama, such as what may be determined to be the most scenic part of the panorama.
  • the foreground location may be selected to cover points of higher contrast along the join seam for the background regions.
  • face analysis techniques and methods may advantageously be employed to further enhance the appearance of such foreground silhouettes, such as those described US patent 7,565,030 or US published applications 2009/0179998, 2008/0013798, 2008/0292193, 2008/0175481, 2008/0266419, 2009/0196466, 2009/0244296, 2009/0190803, 2008/0205712, 2008/0219517, and 2009/0185753, and published PCT app WO2007/142621, and US applications serial nos. 12/512,796 and 12/374,040.
  • foreground facial regions After separating foreground facial regions, these may be enhanced by applying a variety of image processing techniques, prior to being recomposited onto the background panorama scene.
  • the foreground /background separation and independent analyzing and processing of the foreground and background image components in accordance with these embodiments may be used to advantageously provide enhanced panorama images.
  • Figure 13 further illustrates an image blending process in accordance with certain
  • FIG. 12 Five images are shown in Figure 12 to be joined to form a panorama image slightly less than five times wider than any of the component images.
  • Image blending rectangles are illustrated below where the seams will be formed when the panorama image is formed.
  • a portion of the formed panorama image is illustrated below the image blending rectangles and component images in Figure 13.
  • selection of a surface can be used to indicate to a panoramic user interface which projection algorithm is to be applied.
  • a 3D projection algorithm can be used, for example, in accordance with "Image Alignment and Stitching: A tutorial” by Richard Szeliski (Preliminary draft, Sep. 27, 2004, Technical Report, MSR-TR-2004-92. pages 8-10) and/or in any of U.S. published patent applications 2006/0062487,
  • Another embodiment also provides a way to create a hi-res (high resolution) panorama using only lo-res (relatively low resolution) images.
  • lo-res is meant here to be interpreted in comparison to the term "hi-res," wherein the low-res image has a lower resolution than the hi-res image, while no particular resolutions are meant.
  • Such embodiment is suitable for lo-res (& low cost) devices. It is an approach that can be combined with camera-on-chip solutions that may compete with HD devices with superior optics & CPU power.
  • a low-resolution panned video sequence of a panoramic scene is acquired. No high-res images are acquired and thus the seam-line is not transferred to a high- res image as in other embodiments. Also, during the sorting process, fewer image frames are discarded. Instead a larger proportion of image frames may be sorted into one of several different panorama image sets.
  • two or more composite panoramic images may be created by performing edge registration and joining of overlapping images.
  • image frames are joined as follows: frames 1+ 4 + 7 + 10 are used to form one panorama image set, then frames 2 + 5 + 8 + 11 form a second and 3 + 6 + 9 + 12 a third; these three distinct image sets are used to create three distinct panoramic images PI, P2, P3 which may be mostly, but not necessarily entirely overlapping. Note that it is not necessary to retain all images; for example there may have been additional images acquired between frames 3 and 4 (say images 3-2, 3-3 and 3-4) but these may have been discarded because there was not sufficient displacement between them, or because PI, P2 and P3 were deemed sufficient, or otherwise.
  • Super-resolution techniques are then applied to create a single high-resolution panoramic image. Edge registration of each panorama frame-set may be assisted using overlapping images from the other frame-sets. PANORAMA VIEWING MODE FOR HANDHELD DEVICE
  • the screen of a camera-phone may open with a full-height view of a central portion of the panorama image (section in brackets in the following illustration):
  • a modification of "keyhole” or “peephole” viewing is provided when the system stores data relating to the motion which was used to capture the original panorama images.
  • This motion data may be used to control the subsequent user viewing of the image.
  • the motion may be measured during capture and during playback, although handheld motion may be simulated on a desktop computer using a mouse input, for example.
  • An alternative embodiment involves playing back the original video clip but controlling the direction and/or speed of playback according to how the user moves the handheld device during viewing.

Landscapes

  • Engineering & Computer Science (AREA)
  • Multimedia (AREA)
  • Signal Processing (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Computing Systems (AREA)
  • Theoretical Computer Science (AREA)
  • Studio Devices (AREA)

Abstract

Une technique permettant de produire une image panoramique consiste à acquérir un ensemble composé d'au moins deux cadres d'image principaux correspondant à des parties d'une scène qui se chevauchent. Une carte ou d'autres informations en rapport avec la réunion desdits cadres d'image sont stockées. Une image panoramique principale est formée par la réunion des cadres d'image principaux sur la base de la carte ou des informations obtenues grâce à un processus à faible résolution similaire.
PCT/EP2010/064107 2009-12-11 2010-09-24 Imagerie panoramique WO2011069698A1 (fr)

Priority Applications (1)

Application Number Priority Date Filing Date Title
EP10762898.4A EP2545411B1 (fr) 2009-12-11 2010-09-24 Imagerie panoramique

Applications Claiming Priority (10)

Application Number Priority Date Filing Date Title
US12/636,647 2009-12-11
US12/636,639 US20110141225A1 (en) 2009-12-11 2009-12-11 Panorama Imaging Based on Low-Res Images
US12/636,629 2009-12-11
US12/636,608 2009-12-11
US12/636,618 US8294748B2 (en) 2009-12-11 2009-12-11 Panorama imaging using a blending map
US12/636,647 US20110141226A1 (en) 2009-12-11 2009-12-11 Panorama imaging based on a lo-res map
US12/636,629 US20110141229A1 (en) 2009-12-11 2009-12-11 Panorama imaging using super-resolution
US12/636,639 2009-12-11
US12/636,618 2009-12-11
US12/636,608 US20110141224A1 (en) 2009-12-11 2009-12-11 Panorama Imaging Using Lo-Res Images

Publications (1)

Publication Number Publication Date
WO2011069698A1 true WO2011069698A1 (fr) 2011-06-16

Family

ID=43304654

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/EP2010/064107 WO2011069698A1 (fr) 2009-12-11 2010-09-24 Imagerie panoramique

Country Status (2)

Country Link
EP (1) EP2545411B1 (fr)
WO (1) WO2011069698A1 (fr)

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8294748B2 (en) 2009-12-11 2012-10-23 DigitalOptics Corporation Europe Limited Panorama imaging using a blending map
WO2014004380A1 (fr) * 2012-06-25 2014-01-03 Google Inc. Réglages du niveau de détail basés sur le mouvement
WO2014082921A2 (fr) * 2012-11-27 2014-06-05 DigitalOptics Corporation Europe Limited Dispositif de prise de vue numérique à mode panorama
US9118833B2 (en) 2010-11-29 2015-08-25 Fotonation Limited Portrait image synthesis from multiple images captured on a handheld device
EP2983136A2 (fr) 2013-03-18 2016-02-10 FotoNation Limited Procédé et appareil d'estimation de mouvement
US10080006B2 (en) 2009-12-11 2018-09-18 Fotonation Limited Stereoscopic (3D) panorama creation on handheld device
GB2578966A (en) * 2018-10-23 2020-06-03 Hensoldt Optronics Gmbh Method for displaying a panoramic image on an image reproduction device
CN114072801A (zh) * 2020-06-05 2022-02-18 Zillow公司 用于建筑位置的全景图像在移动设备上的自动生成和随后使用

Families Citing this family (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US9589362B2 (en) 2014-07-01 2017-03-07 Qualcomm Incorporated System and method of three-dimensional model generation
US9607388B2 (en) 2014-09-19 2017-03-28 Qualcomm Incorporated System and method of pose estimation
US9911242B2 (en) 2015-05-14 2018-03-06 Qualcomm Incorporated Three-dimensional model generation
US10373366B2 (en) 2015-05-14 2019-08-06 Qualcomm Incorporated Three-dimensional model generation
US10304203B2 (en) 2015-05-14 2019-05-28 Qualcomm Incorporated Three-dimensional model generation
US9936128B2 (en) 2015-05-20 2018-04-03 Google Llc Automatic detection of panoramic gestures
US10681327B2 (en) 2016-09-20 2020-06-09 Cyberlink Corp. Systems and methods for reducing horizontal misalignment in 360-degree video
US10341568B2 (en) 2016-10-10 2019-07-02 Qualcomm Incorporated User interface to assist three dimensional scanning of objects

Citations (39)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5982951A (en) * 1996-05-28 1999-11-09 Canon Kabushiki Kaisha Apparatus and method for combining a plurality of images
US20040130626A1 (en) 2002-10-15 2004-07-08 Makoto Ouchi Panoramic composition of multiple image data
US20040189849A1 (en) * 2003-03-31 2004-09-30 Hofer Gregory V. Panoramic sequence guide
US20050008254A1 (en) 2003-04-15 2005-01-13 Makoto Ouchi Image generation from plurality of images
US20050200706A1 (en) 2003-10-14 2005-09-15 Makoto Ouchi Generation of static image data from multiple image data
US20060018547A1 (en) 2003-11-27 2006-01-26 Makoto Ouchi Image processing device and a method for the same
US20060062487A1 (en) 2002-10-15 2006-03-23 Makoto Ouchi Panorama synthesis processing of a plurality of image data
US20060171464A1 (en) 2005-02-03 2006-08-03 Samsung Electronics Co., Ltd. Method and apparatus for motion estimation
US20060182437A1 (en) 2005-02-11 2006-08-17 Williams Karen E Method and apparatus for previewing a panoramic image on a digital camera
US20060268130A1 (en) 2005-05-26 2006-11-30 Williams Karen E In-camera panorama stitching method and apparatus
US20060285754A1 (en) 2004-08-16 2006-12-21 Eran Steinberg Indoor/Outdoor Classification in Digital Images
US20070081081A1 (en) * 2005-10-07 2007-04-12 Cheng Brett A Automated multi-frame image capture for panorama stitching using motion sensor
US20070147820A1 (en) 2005-12-27 2007-06-28 Eran Steinberg Digital image acquisition system with portrait mode
US20070269108A1 (en) 2006-05-03 2007-11-22 Fotonation Vision Limited Foreground / Background Separation in Digital Images
WO2007142621A1 (fr) 2006-06-02 2007-12-13 Fotonation Vision Limited Modification de paramètres de post-visualisation pour des images numériques utilisant des informations de caractéristiques ou de région d'image
US7315631B1 (en) 2006-08-11 2008-01-01 Fotonation Vision Limited Real-time face tracking in a digital image acquisition device
US20080013798A1 (en) 2006-06-12 2008-01-17 Fotonation Vision Limited Advances in extending the aam techniques from grayscale to color images
US7403643B2 (en) 2006-08-11 2008-07-22 Fotonation Vision Limited Real-time face tracking in a digital image acquisition device
US20080175481A1 (en) 2007-01-18 2008-07-24 Stefan Petrescu Color Segmentation
US20080205712A1 (en) 2007-02-28 2008-08-28 Fotonation Vision Limited Separating Directional Lighting Variability in Statistical Face Modelling Based on Texture Space Decomposition
US20080219517A1 (en) 2007-03-05 2008-09-11 Fotonation Vision Limited Illumination Detection Using Classifier Chains
US20080219581A1 (en) 2007-03-05 2008-09-11 Fotonation Vision Limited Image Processing Method and Apparatus
US7440593B1 (en) 2003-06-26 2008-10-21 Fotonation Vision Limited Method of improving orientation and color balance of digital images using face detection information
US20080266419A1 (en) 2007-04-30 2008-10-30 Fotonation Ireland Limited Method and apparatus for automatically controlling the decisive moment for an image acquisition device
US20080292193A1 (en) 2007-05-24 2008-11-27 Fotonation Vision Limited Image Processing Method and Apparatus
US20080309769A1 (en) 2007-06-14 2008-12-18 Fotonation Ireland Limited Fast Motion Estimation Method
US7469071B2 (en) 2006-02-14 2008-12-23 Fotonation Vision Limited Image blurring
US20090003652A1 (en) 2006-08-11 2009-01-01 Fotonation Ireland Limited Real-time face tracking with reference images
US20090021576A1 (en) 2007-07-18 2009-01-22 Samsung Electronics Co., Ltd. Panoramic image production
US20090022422A1 (en) 2007-07-18 2009-01-22 Samsung Electronics Co., Ltd. Method for constructing a composite image
US20090080713A1 (en) 2007-09-26 2009-03-26 Fotonation Vision Limited Face tracking in a camera processor
US20090179998A1 (en) 2003-06-26 2009-07-16 Fotonation Vision Limited Modification of Post-Viewing Parameters for Digital Images Using Image Region or Feature Information
US7565030B2 (en) 2003-06-26 2009-07-21 Fotonation Vision Limited Detecting orientation of digital images using face detection information
US20090185753A1 (en) 2008-01-18 2009-07-23 Fotonation Ireland Limited Image processing method and apparatus
US20090190803A1 (en) 2008-01-29 2009-07-30 Fotonation Ireland Limited Detecting facial expressions in digital images
US20090196466A1 (en) 2008-02-05 2009-08-06 Fotonation Vision Limited Face Detection in Mid-Shot Digital Images
US20090244296A1 (en) 2008-03-26 2009-10-01 Fotonation Ireland Limited Method of making a digital camera image of a scene including the camera user
US7606417B2 (en) 2004-08-16 2009-10-20 Fotonation Vision Limited Foreground/background segmentation in digital images with differential exposure calculations
US20090273685A1 (en) 2006-02-14 2009-11-05 Fotonation Vision Limited Foreground/Background Segmentation in Digital Images

Patent Citations (44)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5982951A (en) * 1996-05-28 1999-11-09 Canon Kabushiki Kaisha Apparatus and method for combining a plurality of images
US20060062487A1 (en) 2002-10-15 2006-03-23 Makoto Ouchi Panorama synthesis processing of a plurality of image data
US20040130626A1 (en) 2002-10-15 2004-07-08 Makoto Ouchi Panoramic composition of multiple image data
US20040189849A1 (en) * 2003-03-31 2004-09-30 Hofer Gregory V. Panoramic sequence guide
US20050008254A1 (en) 2003-04-15 2005-01-13 Makoto Ouchi Image generation from plurality of images
US7440593B1 (en) 2003-06-26 2008-10-21 Fotonation Vision Limited Method of improving orientation and color balance of digital images using face detection information
US7565030B2 (en) 2003-06-26 2009-07-21 Fotonation Vision Limited Detecting orientation of digital images using face detection information
US20090179998A1 (en) 2003-06-26 2009-07-16 Fotonation Vision Limited Modification of Post-Viewing Parameters for Digital Images Using Image Region or Feature Information
US20050200706A1 (en) 2003-10-14 2005-09-15 Makoto Ouchi Generation of static image data from multiple image data
US20060018547A1 (en) 2003-11-27 2006-01-26 Makoto Ouchi Image processing device and a method for the same
US20060285754A1 (en) 2004-08-16 2006-12-21 Eran Steinberg Indoor/Outdoor Classification in Digital Images
US7606417B2 (en) 2004-08-16 2009-10-20 Fotonation Vision Limited Foreground/background segmentation in digital images with differential exposure calculations
US20060171464A1 (en) 2005-02-03 2006-08-03 Samsung Electronics Co., Ltd. Method and apparatus for motion estimation
US20060182437A1 (en) 2005-02-11 2006-08-17 Williams Karen E Method and apparatus for previewing a panoramic image on a digital camera
US20060268130A1 (en) 2005-05-26 2006-11-30 Williams Karen E In-camera panorama stitching method and apparatus
US20070081081A1 (en) * 2005-10-07 2007-04-12 Cheng Brett A Automated multi-frame image capture for panorama stitching using motion sensor
US20070147820A1 (en) 2005-12-27 2007-06-28 Eran Steinberg Digital image acquisition system with portrait mode
US7469071B2 (en) 2006-02-14 2008-12-23 Fotonation Vision Limited Image blurring
US20090040342A1 (en) 2006-02-14 2009-02-12 Fotonation Vision Limited Image Blurring
US20090273685A1 (en) 2006-02-14 2009-11-05 Fotonation Vision Limited Foreground/Background Segmentation in Digital Images
US20070269108A1 (en) 2006-05-03 2007-11-22 Fotonation Vision Limited Foreground / Background Separation in Digital Images
WO2007142621A1 (fr) 2006-06-02 2007-12-13 Fotonation Vision Limited Modification de paramètres de post-visualisation pour des images numériques utilisant des informations de caractéristiques ou de région d'image
US20080013798A1 (en) 2006-06-12 2008-01-17 Fotonation Vision Limited Advances in extending the aam techniques from grayscale to color images
US20090003652A1 (en) 2006-08-11 2009-01-01 Fotonation Ireland Limited Real-time face tracking with reference images
US20090263022A1 (en) 2006-08-11 2009-10-22 Fotonation Vision Limited Real-Time Face Tracking in a Digital Image Acquisition Device
US7315631B1 (en) 2006-08-11 2008-01-01 Fotonation Vision Limited Real-time face tracking in a digital image acquisition device
US7460695B2 (en) 2006-08-11 2008-12-02 Fotonation Vision Limited Real-time face tracking in a digital image acquisition device
US7469055B2 (en) 2006-08-11 2008-12-23 Fotonation Vision Limited Real-time face tracking in a digital image acquisition device
US7403643B2 (en) 2006-08-11 2008-07-22 Fotonation Vision Limited Real-time face tracking in a digital image acquisition device
US7460694B2 (en) 2006-08-11 2008-12-02 Fotonation Vision Limited Real-time face tracking in a digital image acquisition device
US20080175481A1 (en) 2007-01-18 2008-07-24 Stefan Petrescu Color Segmentation
US20080205712A1 (en) 2007-02-28 2008-08-28 Fotonation Vision Limited Separating Directional Lighting Variability in Statistical Face Modelling Based on Texture Space Decomposition
US20080219581A1 (en) 2007-03-05 2008-09-11 Fotonation Vision Limited Image Processing Method and Apparatus
US20080219517A1 (en) 2007-03-05 2008-09-11 Fotonation Vision Limited Illumination Detection Using Classifier Chains
US20080266419A1 (en) 2007-04-30 2008-10-30 Fotonation Ireland Limited Method and apparatus for automatically controlling the decisive moment for an image acquisition device
US20080292193A1 (en) 2007-05-24 2008-11-27 Fotonation Vision Limited Image Processing Method and Apparatus
US20080309769A1 (en) 2007-06-14 2008-12-18 Fotonation Ireland Limited Fast Motion Estimation Method
US20090022422A1 (en) 2007-07-18 2009-01-22 Samsung Electronics Co., Ltd. Method for constructing a composite image
US20090021576A1 (en) 2007-07-18 2009-01-22 Samsung Electronics Co., Ltd. Panoramic image production
US20090080713A1 (en) 2007-09-26 2009-03-26 Fotonation Vision Limited Face tracking in a camera processor
US20090185753A1 (en) 2008-01-18 2009-07-23 Fotonation Ireland Limited Image processing method and apparatus
US20090190803A1 (en) 2008-01-29 2009-07-30 Fotonation Ireland Limited Detecting facial expressions in digital images
US20090196466A1 (en) 2008-02-05 2009-08-06 Fotonation Vision Limited Face Detection in Mid-Shot Digital Images
US20090244296A1 (en) 2008-03-26 2009-10-01 Fotonation Ireland Limited Method of making a digital camera image of a scene including the camera user

Non-Patent Citations (4)

* Cited by examiner, † Cited by third party
Title
DIJKSTRA, E. W.: "A note on two problems in connexion with graphs", NUMERISCHE MATHEMATIK, vol. 1, 1959, pages 269 - 271
RICHARD SZELISKI: "Image Alignment and Stitching: A Tutorial", TECHNICAL REPORT, MSR-TR-2004-92, 27 September 2004 (2004-09-27), pages 8 - 10
SETHIAN, A FAST MARCHING LEVEL SET METHOD FOR MONOTONICALLY ADVANCING FRONTS, PROCEEDINGS FROM THE NATIONAL ACADEMY OF SCIENCES, 20 October 1995 (1995-10-20)
SNIEDOVICH, M.: "Dijkstra's algorithm revisited: the dynamic programming connexion", (PDF). JOURNAL OF CONTROL AND CYBERNETICS, vol. 35, no. 3, 2006, pages 599 - 620

Cited By (20)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US10080006B2 (en) 2009-12-11 2018-09-18 Fotonation Limited Stereoscopic (3D) panorama creation on handheld device
US8294748B2 (en) 2009-12-11 2012-10-23 DigitalOptics Corporation Europe Limited Panorama imaging using a blending map
US11115638B2 (en) 2009-12-11 2021-09-07 Fotonation Limited Stereoscopic (3D) panorama creation on handheld device
US9118833B2 (en) 2010-11-29 2015-08-25 Fotonation Limited Portrait image synthesis from multiple images captured on a handheld device
US9456128B2 (en) 2010-11-29 2016-09-27 Fotonation Limited Portrait image synthesis from multiple images captured on a handheld device
WO2014004380A1 (fr) * 2012-06-25 2014-01-03 Google Inc. Réglages du niveau de détail basés sur le mouvement
US9153011B2 (en) 2012-06-25 2015-10-06 Google Inc. Movement based level of detail adjustments
US11070723B2 (en) 2012-11-27 2021-07-20 Fotonation Limited Digital image capture device having a panorama mode
WO2014082921A2 (fr) * 2012-11-27 2014-06-05 DigitalOptics Corporation Europe Limited Dispositif de prise de vue numérique à mode panorama
WO2014082921A3 (fr) * 2012-11-27 2014-07-24 DigitalOptics Corporation Europe Limited Dispositif de prise de vue numérique à mode panorama
US11910093B2 (en) 2012-11-27 2024-02-20 Adeia Imaging Llc Digital image capture device having a panorama mode
US9736370B2 (en) 2012-11-27 2017-08-15 Fotonation Limited Digital image capture device having a panorama mode
US10229504B2 (en) 2013-03-18 2019-03-12 Fotonation Limited Method and apparatus for motion estimation
EP2983136A2 (fr) 2013-03-18 2016-02-10 FotoNation Limited Procédé et appareil d'estimation de mouvement
GB2578966A (en) * 2018-10-23 2020-06-03 Hensoldt Optronics Gmbh Method for displaying a panoramic image on an image reproduction device
GB2578966B (en) * 2018-10-23 2022-11-30 Hensoldt Optronics Gmbh Method for displaying a panoramic image on an image reproduction device
CN114072801A (zh) * 2020-06-05 2022-02-18 Zillow公司 用于建筑位置的全景图像在移动设备上的自动生成和随后使用
EP3966725A4 (fr) * 2020-06-05 2022-06-29 Zillow, Inc. Génération automatisée, sur des dispositifs mobiles, d'images panoramiques pour des emplacements de bâtiment et leur utilisation ultérieure
AU2021282989B2 (en) * 2020-06-05 2023-04-13 MFTB Holdco, Inc. Automated generation on mobile devices of panorama images for building locations and subsequent use
AU2021282989A9 (en) * 2020-06-05 2023-04-13 MFTB Holdco, Inc. Automated generation on mobile devices of panorama images for building locations and subsequent use

Also Published As

Publication number Publication date
EP2545411B1 (fr) 2014-02-12
EP2545411A1 (fr) 2013-01-16

Similar Documents

Publication Publication Date Title
US11115638B2 (en) Stereoscopic (3D) panorama creation on handheld device
US8294748B2 (en) Panorama imaging using a blending map
EP2545411B1 (fr) Imagerie panoramique
US20110141226A1 (en) Panorama imaging based on a lo-res map
US20110141229A1 (en) Panorama imaging using super-resolution
US20110141225A1 (en) Panorama Imaging Based on Low-Res Images
US20110141224A1 (en) Panorama Imaging Using Lo-Res Images
US20120019613A1 (en) Dynamically Variable Stereo Base for (3D) Panorama Creation on Handheld Device
US20120019614A1 (en) Variable Stereo Base for (3D) Panorama Creation on Handheld Device
US10600157B2 (en) Motion blur simulation
JP5389697B2 (ja) 高品質な合成パノラマ画像のイン・カメラ生成
WO2021073331A1 (fr) Dispositif et procédé d'acquisition d'image floutée au zoom basés sur un dispositif terminal
JP4513906B2 (ja) 画像処理装置、画像処理方法、プログラム及び記録媒体
CN113271415B (zh) 具有多个图像源的图像处理的方法和系统
JP4658223B2 (ja) 画像生成方法、装置、およびそのプログラム並びにプログラムを記録した記録媒体
TW201251447A (en) Positional sensor-assisted image registration for panoramic photography
CN113271406B (zh) 具有多个图像源的图像处理的方法和系统
JP2009118484A (ja) オブジェクト追跡を用いたデジタル画像の手ぶれ補正装置及び方法
JP6656035B2 (ja) 画像処理装置、撮像装置および画像処理装置の制御方法
JP6261205B2 (ja) 画像処理装置
JP2017143354A (ja) 画像処理装置及び画像処理方法
JP2009118434A (ja) ぶれ補正装置及び撮像装置
JP2009042900A (ja) 撮像装置および画像選択装置
JP2019041188A (ja) 画像処理装置、撮像装置、画像処理装置の制御方法およびプログラム
JP7409604B2 (ja) 画像処理装置、撮像装置、画像処理方法、プログラムおよび記録媒体

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 10762898

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

WWE Wipo information: entry into national phase

Ref document number: 2010762898

Country of ref document: EP