WO2015086537A1 - Procédé de construction d'un ensemble de correspondances de couleurs à partir d'un ensemble de correspondances de caractéristiques dans un ensemble d'images correspondantes - Google Patents

Procédé de construction d'un ensemble de correspondances de couleurs à partir d'un ensemble de correspondances de caractéristiques dans un ensemble d'images correspondantes Download PDF

Info

Publication number
WO2015086537A1
WO2015086537A1 PCT/EP2014/076920 EP2014076920W WO2015086537A1 WO 2015086537 A1 WO2015086537 A1 WO 2015086537A1 EP 2014076920 W EP2014076920 W EP 2014076920W WO 2015086537 A1 WO2015086537 A1 WO 2015086537A1
Authority
WO
WIPO (PCT)
Prior art keywords
color
correspondences
feature
correspondence
colors
Prior art date
Application number
PCT/EP2014/076920
Other languages
English (en)
Inventor
Hasan SHEIKH FARIDUL
Jurgen Stauder
Christel Chamaret
Alain Tremeau
Original Assignee
Thomson Licensing
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 EP14306473.1A external-priority patent/EP3001381A1/fr
Application filed by Thomson Licensing filed Critical Thomson Licensing
Publication of WO2015086537A1 publication Critical patent/WO2015086537A1/fr

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T5/00—Image enhancement or restoration
    • G06T5/50—Image enhancement or restoration using two or more images, e.g. averaging or subtraction
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T5/00—Image enhancement or restoration
    • G06T5/77—Retouching; Inpainting; Scratch removal
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/90—Determination of colour characteristics
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10016—Video; Image sequence
    • G06T2207/10021—Stereoscopic video; Stereoscopic image sequence
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10024—Color image
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20004—Adaptive image processing
    • G06T2207/20008—Globally adaptive

Definitions

  • the invention relates to color and feature correspondences between different images or views, and to the compensation of colors between these images.
  • Geometric differences between these views or versions can be caused by parallax in case of stereo images and by cropping, by zoom, by rotation, or by combinations of them, or by other geometric transforms in case of film scans.
  • geometric differences can be caused by object or by camera motion.
  • Color differences are caused for example by shading, by change of illumination direction, color or distribution, by specular reflections, by cast shadows and other by photometric effects.
  • Color differences are also being caused for example by non-calibrated cameras, by non-calibrated film scanners, by automatically changing exposure settings, by automatic white balancing or even by physical light effects in the scene.
  • Color difference compensation is often the first step in image or video signal processing of multiple views or stereoscopic pictures of the same scene as other steps such as disparity estimation or data compression benefited from low color difference.
  • color differences can disturb artistically the produced content, notably when editing involves the temporal reordering of images.
  • color mapping also called color transfer, color compensation or color correction.
  • Such a color mapping has the task of mapping the color channel coordinates of an image to be suitable for further color signal processing, color signal transmission or color reproduction.
  • GFC Geometric Feature Correspondences
  • SIFT Scale Invariant Feature Transformation
  • outliers are either coming from spatial positioning errors of feature points or from an error in the matching of feature points - so called “false positives”.
  • Wrong GFC, outliers are generally not positioned on the same semantic image details or content in both images, meaning the left and the right images in the case of stereo imaging, or in any two or more images in case of scanning, motion picture, video and broadcasting. There are known methods to remove those outliers, using for instance geometric constraints such as a homographic model, used notably in 3D imaging.
  • a set of pairs of corresponding colors is extracted.
  • a color from one image would generally correspond to a color from the other image. Therefore, from this set, a Color Look-Up Table (LUT) can generally be formed that can be used to compensate the colors of one image in order to be closer to the colors of the other image.
  • LUT Color Look-Up Table
  • Hacohen et al. is based on "Generalized PatchMatch" with an iterative tonal and color correction of the input image, aggregation of consistent regions, and locally narrowing search range of matched transformation.
  • GFC are not always very precise. They suffer from low accuracy in feature's location, shape, and orientation. Main reason for the low accuracy of GFC is estimation errors of feature point detectors due to complex scene geometry, occlusions, and challenging scene illumination. For example, when the curvature of surfaces close to a GFC is high, spatial correspondence of pixels in the neighborhood of a GFC cannot easily be derived from the GFC and extracted color correspondences will be noisy. In general, if the GFC are noisy, the extraction of corresponding colors by known methods may fail completely. In other cases, corresponding colors extracted from noisy GFC by known methods may be noisy themselves. Summary of invention
  • a subject of the invention is a method for building a set of color
  • the method comprises also the steps of:
  • a subject of the invention is also a method for compensating color differences between different images comprising the steps of:
  • correspondences according the above method of for building a set of color correspondences, and compensating color differences between said different images by replacing the colors of at least one image by the corresponding colors according to said built set of color correspondences.
  • a subject of the invention is also a method for building a set of color correspondences from a set of feature correspondences between features points in a plurality of different images, the method comprising, for each feature correspondence of said set:
  • the second step of the method above preferably starts with selecting at least three pixels at different spatial positions in the spatial neighborhood of each feature point of said feature correspondence such that the spatial position of any selected pixel in the spatial neighborhood of a feature point of said feature correspondence corresponds spatially to the spatial position of a selected pixel in the spatial neighborhood of any other feature point of said feature correspondence.
  • the method comprises also, as illustrated on figure 1 , a selection, in each provided set of local color correspondences, of color correspondences that fit a color correspondence model according to a defined fitting criterion, to provide a set of local robust color correspondences.
  • This color correspondence model is such that to a color in an image of said plurality belonging to any local color correspondence of the provided set correspond a color in each of the other images of said plurality, resulting in a set of modeled color correspondences.
  • the set of local robust color correspondences is built with all the local color correspondences that are close enough to their corresponding modeled color correspondences, according to defined fitting criteria. Such a method of selection of color correspondences allows advantageously removing outliers independently of geometric features or constraints of the images.
  • the color correspondence model could be for instance the straight line that would approximate the best the local color correspondences of a set local color correspondences in at least one of these color channels. Such a model is then linear in the RGB color space.
  • the color correspondences of this set that do not fit such a linear relationship well enough according to a defined criteria are ignored as being "non robust", i.e. are not part of the set of robust local color correspondences.
  • the defined fitting criteria could be for instance, in said at least one of these color channels, a maximum distance from the straight line representing the linear model, or a distance less than a defined threshold between the position of a local color correspondence and the position of the corresponding modeled color correspondence in a multidimensional space, having as many dimension as there are colors in each color correspondence.
  • the method comprises also a selection, in each provided set of local color correspondences and among said local color correspondences of said set, any color correspondence having a color which, after being mapped by a color model into mapped colors have color coordinates in the at least one color channel that are distant less than a defined threshold from the corresponding color coordinates of the other colors of this color correspondence, thus resulting in a set of local robust color correspondences.
  • said color model is based on a linear relationship between the color coordinates in the said at least one color channel of a color and the color coordinates in the same at least one color channel of said mapped colors.
  • a subject of the invention is also a method for compensating color differences between different images comprising:
  • a subject of the invention is also an apparatus for building a set of color correspondences in a plurality of different images, comprising:
  • a feature correspondences building module for building a set of feature correspondences between features points in the plurality of different images received by said receiving module
  • a warping module configured for warping the spatial neighborhood of at least one of the corresponding feature points of a feature correspondence built by said feature correspondences building module such that, after warping, the spatial positions of pixels in the neighborhoods of the feature points
  • a generating module for generating a set of at least three local color
  • correspondences by bringing in correspondence colors of pixels having spatial positions that are brought in spatial correspondence by the warping module in the neighborhoods of the feature points of said feature correspondence.
  • the apparatus comprises also a denoising module configured for selecting, in a set of local color correspondences generated by said generating module, color correspondences that fit a color correspondence model according to a defined fitting criterion, resulting in a set of local robust color correspondences.
  • a denoising module configured for selecting, in a set of local color correspondences generated by said generating module, color correspondences that fit a color correspondence model according to a defined fitting criterion, resulting in a set of local robust color correspondences.
  • a subject of the invention is also a system for compensating color differences between different images comprising:
  • a color compensation module configured for compensating color differences between the different images of said plurality by replacing the colors of at least one image of said plurality by their corresponding colors according to the set of color correspondences.
  • FIG. 1 illustrates a main embodiment of the method of building a set of local robust color correspondence according to the invention
  • FIG. 2 illustrates an example of implementation of the method according to the invention in case of two images, a first image, or reference view, and a second image, or test view, with geometrically normalized neighbourhoods of a feature correspondence from first and second images l L , l R with low spatial accuracy (presence of horizontal shift) marked by ellipses, with two features correspondences F : L ⁇ -> F : R and F 2 L ⁇ -> F 2 R in these two images;
  • FIG. 3 illustrates one set of local color correspondences obtained for one of the feature correspondences illustrated on figure 2;
  • FIG. 4 black pixels corresponds to color correspondences which do not follow a linear model(s), that are not taken into account in color correspondences selected on figure 5;
  • Figure 5 illustrates the color correspondences that are selected from the color correspondences represented on figure 3 using robust linear fitting based on iteratively re-weighted least squares.
  • the steps of the various elements of the invention may be provided through the use of dedicated hardware as well as hardware capable of executing software in association with appropriate software.
  • the hardware may notably include, without limitation, digital signal processor (“DSP”) hardware, read-only memory (“ROM”) for storing software, random access memory (“RAM”), and non-volatile storage.
  • DSP digital signal processor
  • ROM read-only memory
  • RAM random access memory
  • non-volatile storage non-volatile storage.
  • a receiving module for receiving a pair of a first and a second images
  • a feature correspondences building module for building a set of feature correspondences between feature points of the first image and feature points of the second image
  • a warping module configured for geometric normalization and to bring pixels of the neighborhood of any feature point of the first image into spatial
  • a generating module for generating a set of at least three color
  • a denoising module configured for selecting, in a set of local color
  • a color compensation module configured for compensating color differences between the two images of the pair by replacing the colors of an image of this pair by their corresponding colors according to the set of robust color
  • the invention proposes a method to calculate a set of pairs of corresponding colors, each pair having one color from a first image and a corresponding color from the second image, colors being described by color coordinates, given a set of feature correspondences, each feature correspondence being a pair of feature points, simply called features, each pair having a first feature from the first image and a corresponding second feature in the second image, a feature point being a spatial position in the image and its spatial neighborhood, the method comprising notably, for each feature correspondence between a feature of the first image and a feature of the second image, the following steps:
  • step 5 building a color look up table able to transform color coordinates of colors from the first image into color coordinates of colors of the second image
  • the first step is to generate or build feature correspondences between the first and the second image. Detection of features in these images as well as creating correspondences between these detected features is well-known. For example, Lowe presents in his paper entitled “Distinctive image features from scale- invariant keypoints" published in the. Int'l J. Computer Vision no. 60 pages 91- 1 10 a method called "SIFT" that first detects features in images and then estimates correspondences between the features of two images. Each feature is defined by its position, its scale and its orientation. The position is the spatial coordinates in 2D image space. The scale indicates the size of the region centered around the position that has been used to detect the feature. A large scale indicates a large region and a small scale indicates a small region.
  • the orientation indicates the angular orientation of the feature, between 0 and 360°. Lowe then proposes to describe each feature by a descriptor. Two features from two images are then detected to be a feature correspondence if their descriptors are close. Between the first image and the second image, M feature pi* j?R
  • Figure 2 illustrates two features correspondences F : L ⁇ -> F : R and F 2 L ⁇ -> F 2 R between two images l L and l R .
  • Figure 2 shows an example of the geometrically normalized neighborhood of a feature correspondence for two images l L and l R .
  • the feature neighborhoods are magnified for visualization.
  • figure 2 seems to be a precise good feature correspondence.
  • the accuracy of the feature location is low, as marked by the ellipses in this figure. As explained below, low accuracy in location, scale/shape, orientation of corresponding features creates noisy color correspondences, as shown on figure 3.
  • the second step is collecting M local sets of color correspondences from these M feature correspondences.
  • a feature correspondence as detected above is known to refer to two corresponding image regions, i.e. two corresponding spatial neighborhoods of feature points, wherein each neighborhood includes its feature point and is defined by parameters such as position, scale, and orientation. Given these parameters, the spatial neighborhoods of rJ ⁇ f R- corresponding features are then defined in both images . In ideal case, within the calculated neighborhood of a feature of the first image, each pixel of this neighborhood has a corresponding pixel in the corresponding neighborhood of the corresponding feature of the second image.
  • an additional geometric normalization step is preferably added that is adapted in a manner known per se to match the shape, size and orientation of the corresponding neighborhoods.
  • Such a geometric normalization of corresponding neighborhoods correspond to a warping of at least one of these neighborhoods adapted such that, after warping, these spatial neighborhoods correspond spatially themselves to each other.
  • This warping process is usually an approximation and, after warping, the correspondence between the spatial positions of pixels in the neighborhood of the first feature and the spatial positions of pixels in the neighborhood of the second feature is usually still noisy.
  • Figure 3 illustrates a set of local color correspondences obtained for one of the feature correspondences such as illustrated on figure 2.
  • low accuracy in location, scale/shape, orientation of corresponding features of figure 2 creates noisy color correspondences. That is why the next step is dedicated to the improvement of the "robustness" of the color correspondences in each set.
  • the general approach is to locally process the neighborhood of each feature correspondences before selecting the color correspondences.
  • a locally linear color mapping model is fit to each color channel of the N local color correspondences of the set
  • This local color model is used to improve the robustness of the color correspondences of each set.
  • a local color model is preferably a color mapping model which is linear in the RGB color space.
  • the linear robust fit of the method according to the invention is not limited to three color channels. For example, if there are four color channels, four local models are calculated.
  • the robust fit of the method according to the invention is not limited to linear models.
  • Linear models often fail in modeling the difference of color coordinates between left and right images if the range of color coordinates is large. It is a known method to approximate non-linear functions in a local range of values by a linear function. Using a linear color model, it can be expected that a minority of colors having color coordinates very different from the majority of colors will not be well described by a linear model. It can be therefore advantageous to use a non-linear model such as an exponential model, a quadratic model, a model based on splines or any other type of non-linear function. Since the model is fitted to a local set of color correspondences with is usually limited in number, it is advantageous to use a non-linear color model with a limited number of parameters.
  • n color correspondences of the set of local color correspondences is a nx2 matrix, where the first column is all ones and the second column is composed of the color coordinates ⁇ ⁇ (e.g., R, or G, or B,) of the corresponding colors from image li_ according to
  • ⁇ is a 2 x 1 vector representing the estimated model parameters of the local, linear color model of the color coordinates of this channel.
  • Another possibility is to choose the threshold according to the robust standard deviation of residuals such as proposed by Motulsky and Brown in their paper entitled “Detecting outliers when fitting data with nonlinear regression - a new method based on robust nonlinear regression and the false discovery rate" published in BMC bioinformatics 7(1 ) in 2006.
  • Another possibility is to apply a threshold T in one step to all color channels. For example in case of three color channels red, green and blue, the 3D distance in R,G,B color space between (Rf, Gf, Bf) and the models ⁇ g R + h R R j , g G + h G Gf, g B + h B B ) is compared to a single threshold T. The resulting robustly selected color correspondences form a set of local, robust color correspondences. Other thresholds can be used.
  • Figure 5 illustrates the color coordinates of color correspondences that are selected from the color correspondences represented on figure 3 using the above robust linear fitting based on iteratively re-weighted least squares.
  • a comparison between figures 3 and 5 clearly shows that the color coordinates of the selected color correspondences shown in figure 5 are far closer to linear models than those of figure 3. Pixels having colors not following these linear model(s) are shown as black pixels in figure 4 and are not taken into account in the selected color correspondences shown on figure 5.
  • this color look up table is applied to each color of the first image of the pair, resulting in a color compensated first image.
  • the method according to the invention has the following advantages over known methods:
  • the invention may be implemented in various forms of hardware, software, firmware, special purpose processors, or combinations thereof.
  • the invention may be notably implemented as a combination of hardware and software.
  • the software may be implemented as an application program tangibly embodied on a program storage unit.
  • the application program may be uploaded to, and executed by, a machine comprising any suitable architecture.
  • the machine is implemented on a computer platform having hardware such as one or more central processing units (“CPU"), a random access memory (“RAM”), and input/output (“I/O”) interfaces.
  • CPU central processing units
  • RAM random access memory
  • I/O input/output
  • the computer platform may also include an operating system and microinstruction code.
  • various processes and functions described herein may be either part of the microinstruction code or part of the application program, or any combination thereof, which may be executed by a CPU.
  • various other peripheral units may be connected to the computer platform such as an additional data storage unit and a printing unit.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Image Analysis (AREA)
  • Image Processing (AREA)

Abstract

L'invention porte sur un procédé qui consiste, pour chaque correspondance de caractéristique : à déformer le voisinage spatial d'un point caractéristique de sorte que, après la déformation, les voisinages spatiaux des points caractéristiques, correspondant les uns aux autres dans ladite correspondance de caractéristiques dans les différentes images, correspondent spatialement eux-mêmes les uns aux autres ; à sélectionner au moins trois premières couleurs à trois positions spatiales différentes du voisinage spatial dans une première image et une second couleur dans le voisinage spatiale correspondant d'une autre image différente, ce qui permet d'obtenir un ensemble d'au moins trois correspondances de couleur locales, chaque correspondance de couleur ayant une première couleur dans ladite première image et une seconde couleur dans l'autre image.
PCT/EP2014/076920 2013-12-10 2014-12-08 Procédé de construction d'un ensemble de correspondances de couleurs à partir d'un ensemble de correspondances de caractéristiques dans un ensemble d'images correspondantes WO2015086537A1 (fr)

Applications Claiming Priority (4)

Application Number Priority Date Filing Date Title
EP13306691 2013-12-10
EP13306691.0 2013-12-10
EP14306473.1 2014-09-24
EP14306473.1A EP3001381A1 (fr) 2014-09-24 2014-09-24 Procédé de construction d'un ensemble de correspondances de couleur à partir d'un ensemble de correspondances de caractéristiques dans un ensemble d'images correspondantes

Publications (1)

Publication Number Publication Date
WO2015086537A1 true WO2015086537A1 (fr) 2015-06-18

Family

ID=52101308

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/EP2014/076920 WO2015086537A1 (fr) 2013-12-10 2014-12-08 Procédé de construction d'un ensemble de correspondances de couleurs à partir d'un ensemble de correspondances de caractéristiques dans un ensemble d'images correspondantes

Country Status (1)

Country Link
WO (1) WO2015086537A1 (fr)

Cited By (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US9754237B2 (en) 2015-12-18 2017-09-05 Ricoh Co., Ltd. Index image quality metric
US9805289B2 (en) 2015-12-18 2017-10-31 Ricoh Co., Ltd. Color-based post-processing of images
US9875548B2 (en) 2015-12-18 2018-01-23 Ricoh Co., Ltd. Candidate list generation
US9904990B2 (en) 2015-12-18 2018-02-27 Ricoh Co., Ltd. Single image rectification
US9911213B2 (en) 2015-12-18 2018-03-06 Ricoh Co., Ltd. Panoramic image stitching using objects
US9984451B2 (en) 2015-12-18 2018-05-29 Michael Gormish Linear grouping of recognized items in an image
US10339690B2 (en) 2015-12-18 2019-07-02 Ricoh Co., Ltd. Image recognition scoring visualization
US10514825B2 (en) 2015-12-18 2019-12-24 Ricoh Co., Ltd. Image recognition result visualization over time
US11915431B2 (en) * 2015-12-30 2024-02-27 Texas Instruments Incorporated Feature point identification in sparse optical flow based tracking in a computer vision system

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
FARIDUL HASAN SHEIKH ET AL: "Approximate Cross Channel Color Mapping from Sparse Color Correspondences", 2013 IEEE INTERNATIONAL CONFERENCE ON COMPUTER VISION WORKSHOPS, IEEE, 2 December 2013 (2013-12-02), pages 860 - 867, XP032575686, DOI: 10.1109/ICCVW.2013.118 *
HASAN S F ET AL: "Robust Color Correction for Stereo", VISUAL MEDIA PRODUCTION (CVMP), 2011 CONFERENCE FOR, IEEE, 16 November 2011 (2011-11-16), pages 101 - 108, XP032074523, ISBN: 978-1-4673-0117-6, DOI: 10.1109/CVMP.2011.18 *
PEI-KUEI TSUNG ET AL: "Hybrid color compensation for virtual view synthesis in multiview video applications", IEEE INTERNATIONAL SYMPOSIUM ON CIRCUITS AND SYSTEMS. ISCAS 2010 - 30 MAY-2 JUNE 2010 - PARIS, FRANCE, IEEE, US, 30 May 2010 (2010-05-30), pages 121 - 124, XP031724240, ISBN: 978-1-4244-5308-5 *

Cited By (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US9754237B2 (en) 2015-12-18 2017-09-05 Ricoh Co., Ltd. Index image quality metric
US9805289B2 (en) 2015-12-18 2017-10-31 Ricoh Co., Ltd. Color-based post-processing of images
US9875548B2 (en) 2015-12-18 2018-01-23 Ricoh Co., Ltd. Candidate list generation
US9904990B2 (en) 2015-12-18 2018-02-27 Ricoh Co., Ltd. Single image rectification
US9911213B2 (en) 2015-12-18 2018-03-06 Ricoh Co., Ltd. Panoramic image stitching using objects
US9984451B2 (en) 2015-12-18 2018-05-29 Michael Gormish Linear grouping of recognized items in an image
US10339690B2 (en) 2015-12-18 2019-07-02 Ricoh Co., Ltd. Image recognition scoring visualization
US10445870B2 (en) 2015-12-18 2019-10-15 Ricoh Company, Ltd. Linear grouping of recognized items in an image
US10489893B2 (en) 2015-12-18 2019-11-26 Ricoh Company, Ltd. Single image rectification
US10514825B2 (en) 2015-12-18 2019-12-24 Ricoh Co., Ltd. Image recognition result visualization over time
US11915431B2 (en) * 2015-12-30 2024-02-27 Texas Instruments Incorporated Feature point identification in sparse optical flow based tracking in a computer vision system

Similar Documents

Publication Publication Date Title
US10540806B2 (en) Systems and methods for depth-assisted perspective distortion correction
WO2015086537A1 (fr) Procédé de construction d'un ensemble de correspondances de couleurs à partir d'un ensemble de correspondances de caractéristiques dans un ensemble d'images correspondantes
US9300946B2 (en) System and method for generating a depth map and fusing images from a camera array
EP3158532B1 (fr) Égalisation d'histogramme adaptative locale
CN107077743B (zh) 用于阵列相机的动态校准的系统和方法
JP6663652B2 (ja) ステレオソース映像の補正方法及びその装置
WO2017076106A1 (fr) Procédé et dispositif d'assemblage d'images
US20150124059A1 (en) Multi-frame image calibrator
US11282232B2 (en) Camera calibration using depth data
JP2015197745A (ja) 画像処理装置、撮像装置、画像処理方法及びプログラム
JP2010045613A (ja) 画像識別方法および撮像装置
AU2017251725A1 (en) Calibration of projection systems
JP7024736B2 (ja) 画像処理装置、および画像処理方法、並びにプログラム
CN110909750A (zh) 图像差异检测方法及装置、存储介质、终端
CN110598795A (zh) 图像差异检测方法及装置、存储介质、终端
US10116919B2 (en) Method and arrangement for estimating at least one cross-channel colour mapping model from an set of tuples of corresponding colours relative to at least two images
Hahne et al. PlenoptiCam v1. 0: A light-field imaging framework
Benveniste et al. Nary coded structured light-based range scanners using color invariants
WO2013164043A1 (fr) Procédé et système de détermination d'un modèle de mise en correspondance des couleurs permettant de transformer des couleurs d'une première vue en des couleurs d'au moins une seconde vue
CN117333367A (zh) 基于图像局部特征的图像拼接方法、系统、介质及装置
Sheng et al. Guided colorization using mono-color image pairs
EP3001381A1 (fr) Procédé de construction d'un ensemble de correspondances de couleur à partir d'un ensemble de correspondances de caractéristiques dans un ensemble d'images correspondantes
Ballabeni et al. Intensity histogram equalisation, a colour-to-grey conversion strategy improving photogrammetric reconstruction of urban architectural heritage
Guthier et al. Histogram-based image registration for real-time high dynamic range videos
AU2019201825A1 (en) Multi-scale alignment pattern

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: 14812181

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 14812181

Country of ref document: EP

Kind code of ref document: A1