Classifications

• • G—PHYSICS
  • G06—COMPUTING; CALCULATING OR COUNTING
  • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
  • G06V20/00—Scenes; Scene-specific elements
  • G06V20/35—Categorising the entire scene, e.g. birthday party or wedding scene
• • G—PHYSICS
  • G06—COMPUTING; CALCULATING OR COUNTING
  • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
  • G06V10/00—Arrangements for image or video recognition or understanding
  • G06V10/40—Extraction of image or video features
  • G06V10/56—Extraction of image or video features relating to colour

Definitions

• the object of the present invention relates to the technical field of the image in the general sense and it relates more precisely to the field of the classification of color images according to the place of shooting of the image, that is to say - say either outdoors or indoors.
• the object of the invention finds a particularly advantageous but non-limiting application in the field of macro-segmentation of video images.
• the macro-segmentation of a video aims to classify the shots into semantic units which are the scenes.
• This macro-segmentation of the video in shots is based on the analysis of a signal obtained from successive images of the video. An image reflecting the content of a plan is then chosen as being the representative image thereof.
• the classification of shots according to their location facilitates the segmentation of a video into scenes. Indeed, the internal term is an indication of place, fixed at the time of the first stage of the writing of the scenario, just before the description of the scenes. This term is in opposition to the external term. This indication of the location therefore allows the technical team to know in advance all the scenes that must be filmed indoors or outdoors. The proper use of this term also facilitates the work of the director of photography who is thus able to know the scenes requiring different lighting and to consequently develop the appropriate lighting according to the place of the scene, namely outside or inside. Knowledge of this location index is also important when analyzing video images since it constitutes a semantic index.
• the object of the invention therefore aims to satisfy this need by proposing a method for classifying an image in color with a view to determining the place of shooting of the image, that is to say outdoors or indoors.
• the method according to the invention consists: To determine spectral bands of reference colors corresponding to different values of the temperature of an image, making it possible to characterize an image outdoors or an image indoors,
• the method according to the invention consists in determining as spectral bands of reference colors, the spectral bands of red, yellow and green color ordered in increasing order of their temperature.
• Figure 1 is a block diagram of a device for implementing the method according to the invention.
• Figures 2a to 2i are different histograms of spectra to explain the process according to the invention.
• the object of the invention relates to a device 1 making it possible to classify a color image as a function of the place where the image is taken, that is to say either outdoors (in nature) or indoors (a room, a building, or a cave for example).
• the system 1 comprises an image sensor 2 such as for example a video camera connected at the output to a digitizing means 3 making it possible to digitize the images which are defined in a determined spectral space of Colors, for example RGB, known and defined by the CIE (International Lighting Commission).
• an image sensor 2 such as for example a video camera connected at the output to a digitizing means 3 making it possible to digitize the images which are defined in a determined spectral space of Colors, for example RGB, known and defined by the CIE (International Lighting Commission).
• CIE International Lighting Commission
• the output of the digitizing means 3 is connected to a data processing system 4 such as a computer comprising. programmed means allowing in particular to analyze the images in order to determine the place of shooting for each selected image.
• the data processing system 4 is connected to storage means 5 making it possible to record each analyzed image as well as for each of them, the associated location indication index, that is to say an image in indoors or outdoors.
• the system 4 comprises means making it possible to determine spectral bands of reference color Bj corresponding to different values of the temperature of an image and making it possible to characterize an image outdoors or indoors.
• the thermodynamic temperature of a light source can be estimated by an analysis of the spectral distribution M f of the radiation and by a classification of the color associated therewith.
• a spectral distribution M f of a radiation is presented according to seven spectral bands as illustrated in table 1 below:
• the different spectral bands are ordered in descending order of their temperature. It must be considered that artificial light, unlike natural light, has a spectrum whose values are mainly found in the spectral band of red color and, and in a minority in the spectral bands of Green and blue colors. Thus, an image with a predominantly red spectrum whose temperature is considered to be the warmest temperature, corresponds to an interior light (case of the temperature of the interior colors of a room, in a building, in a cave, etc. ). In this case, the image is classified as an indoor image, that is to say, the image was taken inside a building, etc.
• an image with a predominantly green spectrum whose temperature is classified among the hottest temperatures (case of the temperature of the color of the sky, color of the sea, etc.) corresponds to light from the outside.
• the image is classified as an outdoor image, that is to say one whose shooting is carried out in nature.
• spectral bands of reference color the spectral bands of red, yellow and green color ordered in increasing order of their temperature.
• the spectral band of blue color is in certain cases difficult to quantify in a spectrum of light. Indeed, it has been found that cinema producers tend to compensate for the excess of the blue spectrum so that its analysis can distort the results.
• the spectral band of yellow color is easy to quantify insofar as it is present practically in all the spectra. Even if its presence rate in a spectrum depends on the type of light, namely natural or artificial, it remains in most cases sufficiently important to be able to be quantified.
• the data processing system 4 also comprises means making it possible to determine, for at least the reference color spectral bands, the distribution of the color spectra of at least part of the color image digitized and selected to be analyzed.
• the digitized color image defined in the determined RGB color space is transformed into an image of the YIQ color space known per se and defined by the NTSC (National Television System Committee).
• the three color components are respectively; Luminance, In phase and Quatrature phase and materialize the three axes (white-black), (red-cyan) and (magenta-green) of space colours.
• Table 2 The transformation of the RGB spectral space into the YIQ spectral space is given by table 2 below:
• RN NTSC system 1.910 -0.533 -0.288 X primary receivers RN, G » f ⁇ - 0.985 2.000 -0.028 Y
• each elementary block can comprise 16 times 16 pixels.
• the luminance average Y and the average in phase I are calculated. It should be noted that the originality of the axis I of the YIQ system is that it represents the axis (red-cyan ) of the color space while the Y axis represents the luminance along the axis (white-black).
• the sum of the mean in luminance Y and the mean in phase I is selected since it represents the region of the image having the most information concerning the temperature.
• the maximum in luminance Y and in phase I in the image corresponds to the region of the image having the highest temperature associated with the light source.
• the method then consists in studying the temperature for the selected block of pixels.
• the spectral distribution Mf obtained by the above formula is described as a function of the seven visible spectral bands presented in Table 1.
• an amplified version A f of the matrix M f is used by the following formula
• Af (i, j) 490 if 490 ⁇ M f (i, j) ⁇ 560 560 if 560 ⁇ M f (i, j) ⁇ 580 580 if 580 ⁇ M f (i, j) ⁇ 600 600 if 600 ⁇ M f (i, j) ⁇ 700
• the method then consists in applying to the amplified matrix A f , a quantification and classification process which is focused on the three reference spectral bands as explained above, namely the spectra of red color.
• the histogram of the spectrum is calculated according to the three spectral bands of reference color, namely red, yellow and green defined above.
• the method then consists in detecting the two dominant spectral bands, that is to say the two most important peaks in the histogram of the spectrum.
• the comparison of the values of these peaks as well as their position with respect to the temperature axis makes it possible to decide whether the components of the spectrum of the block of pixels selected, tend to be closer to the hot spectra or on the contrary of the Rather cold spectra, to classify the image into an indoor image or an outdoor image.
• FIGS. 2a to 2i illustrate the different cases likely to be encountered during the determination of the two dominant spectra of the histogram of the spectrum of an image.
• FIG. 2a gives the case of a completely cold spectrum, so that the image is classified as an interior image
• FIG. 2b corresponds to a completely hot spectrum corresponding to an image classified as an exterior.
• it is a 100% compound spectrum in the red spectral band (cold temperature)
• Figure 2c shows the case of a predominantly green spectrum since the green spectral band is larger than that of red color. Consequently, the corresponding image is classified as an outdoor image because the temperature of the light is rather warm.
• Figure 2d illustrates the case of a predominantly red spectrum since the red spectral band is larger than that of the green color.
• the corresponding light is rather cold and the associated image is classified as an indoor image.
• FIG. 2g illustrates the case where the peak of green color is greater than that of yellow. The associated light is therefore rather on the side of the hot temperature, so that the corresponding image is classified as an outdoor image.
• Figure 2h illustrates the case where the yellow color peak is greater than that of green color. In this case, the image is classified indoors.
• FIG. 2i corresponds to a spectrum composed at 100% in the yellow spectral band.
• a spectrum generally comes from a backlit image.
• the determination of the matrix of the spectral distribution results in a dominance of yellow light of the block which is identified as belonging to an intense light source (sky or midday sun) visible for example by a window.
• the intensity of white indicates whether the image has a backlight effect or an image from the outside.
• This effect is identified by analyzing the grayscale histogram of the entire image. Analysis of the grayscale histogram makes it possible, as a function of the pixels in the white or dark areas, to classify the image respectively as an outdoor image or an indoor image. The accumulation of a large number of pixels in the intense white areas indicates that this is an outdoor image. Otherwise, the image is classified as an indoor image.
• the object of the invention thus relates to a method for classifying a color image as a function of the thermodynamic temperature of the existing light source when the image is taken.
• the taking into account of this temperature which is estimated by the analysis of the spectral distribution of the radiation makes it possible to classify the image according to the place of its shooting namely, indoors or outdoors.

Landscapes

• Engineering & Computer Science (AREA)
• Physics & Mathematics (AREA)
• General Physics & Mathematics (AREA)
• Multimedia (AREA)
• Theoretical Computer Science (AREA)
• Image Analysis (AREA)
• Spectrometry And Color Measurement (AREA)
• Processing Of Color Television Signals (AREA)
• Color Television Image Signal Generators (AREA)
• Image Processing (AREA)
• Studio Devices (AREA)
• Facsimile Image Signal Circuits (AREA)

Applications Claiming Priority (3)

Publications (1)

Family

ID=8857381

Family Applications (1)

Country Status (6)

Families Citing this family (2)

Family Cites Families (9)

• 2000
  • 2000-12-07 FR FR0015933A patent/FR2817986B1/fr not_active Expired - Fee Related
• 2001
  • 2001-12-07 JP JP2002548678A patent/JP2004524726A/ja not_active Withdrawn
  • 2001-12-07 WO PCT/FR2001/003869 patent/WO2002047029A2/fr not_active Application Discontinuation
  • 2001-12-07 US US10/432,622 patent/US20040071341A1/en not_active Abandoned
  • 2001-12-07 EP EP01999921A patent/EP1348200A2/fr not_active Withdrawn
  • 2001-12-07 AU AU2002217204A patent/AU2002217204A1/en not_active Abandoned

Non-Patent Citations (1)

Also Published As

Similar Documents

Legal Events