LV14207B - Method and device for multi-spectral imaging by means of digital rgb sensor - Google Patents

Abstract

Invention relates to methods of digital colour image processing, in particular - to selection of images corresponding to more than three different spectral bands from a single RGB data set. In the proposed method of multi-spectral imaging, the object is illuminated simultaneously at several spectral bands, and the values of Ri-, Gi- un Bi-signals detected at each i-pixel of the image are identified. Further they are compared mutually and with externally determined signal discrimination level S that allows registering only one or two of the R-, G- and/or B-bands within the spectral sensitivity range of the sensor. In order to increase the number of available spectral images, the S-values are continuously variable up to the highest of all possible signal values of the R, G or B band, with condition that linear photo-response of the RGB sensor is ensured. Depending on the S-value, two situations are analyzed - if two colour band signals are registered simultaneously (i.e. B and G, G and R or B and R), or if the signals are registered only at one colour band - and after logical analysis the spectral interval(-s) of the pixel-registered radiation is identified. Further each spectral image is formed from the pixels that correspond to a particular selected spectral range. A device for multi-spectral imaging to implement this method comprises a multi-spectral light source, objective-equipped digital RGB sensor, RGB data set storage device, convertor that converts the RGB data into a set of spectral intensities in accordance with the selected signal discrimination level, selector of images that selects the images related to each particular spectral band, and the output device, e.g. PC-monitor.

Description

Description of the Invention

Technical field

The invention relates to a method for processing digital color images, in particular to selecting images corresponding to more than three different spectral bands from a single RGB image data array.

Known state of the art

Analysis of multispectral image sets allows the identification of areas of objects in the images with various spectral absorption or emission characteristics useful for expert examination of works of art, satellite monitoring of the Earth's surface, clinical diagnosis of the skin, etc. Digital sensors are often used to produce multispectral images - a two-dimensional array of photographic matrices with filters that pass through different spectral regions, such as a rotating disk with a set of different color filters /E.C. Ruvolo et al., Proc. SPIE, Vol. 7548, 75480A, 2010 / and Acoustic or Liquid Crystal Filters spectrally read with electrical signals /http://www.usgs.gov/science/science.php?term=765/.

Digital RGB sensors incorporate three-color (blue, green, and red) spectral filters, which are used for multispectral imaging in various combinations with external spectral filters / US 7612822 B2; US 2009290124 (A1); JP 2008136251 (A) /. Another way to obtain multi-spectral images is to use multiple light sources emitting in different spectral bands, such as LEDs of different colors / WO 2008093988 (Al) /, for sequential illumination of the subject by capturing its own image in each light spectrum band.

These techniques are practical, but their drawback is the need to take several sequential images of the same subject in different spectral bands. First of all, it is a time consuming process. Second, during the capture, the properties of the object may change, such as its movement, in which case either the erroneous results have to be accepted or the processing algorithm must include additional procedures for image stabilization, which delay the rapid acquisition of information. Third, a large amount of unnecessary information about image details is stored. Fourth, such hardware designs are complex and expensive because they include external filtering devices or sets of spectrally different illumination sources in addition to the RGB sensor lens.

Such drawbacks do not exist in the methodology of multispectral analysis of a single RGB image. Pictures taken with a digital RGB sensor such as CCD or CMOS Zhttp.7 / broadcastengineering.com / hdtv / ccd-cmos // are reproduced in color format, with each pixel or pixel first determined by red (R), green (G) and blue (B) the numerical values Ri, Gjun Bj of the signals recorded in the spectral band, where the index i indicates the pixel number, and then assigning this pixel the corresponding color for the corresponding combination of RjGiBj. The spectral sensitivity of the individual R, G, and B channels is determined by the absorption properties of the three types of filter coatings and the matrix photodetector material itself (e.g., silicon). For series-produced RGB series, the three-band spectral sensitivity curves can be found in the product specification; they are also experimentally measurable. Thus, three spectrally selective sub-images in the R, G, and B bands (corresponding to the light emitted or reflected by the object in the red, green, and blue bands of the spectrum) can be further separated from a single RGB color image data array, minus one another or the like. / D. Kapsokalyvas et al., Proc. SPIE, Vol. 7548, 754808, 2010 /. By fixing a certain level of discrimination of the sensor output signals above any two or one of the R, G, B values at any wavelength in the sensor's spectral sensitivity range, it is possible to logically analyze a further six narrower spectral ranges in the RGB array. Spīgulis et al., Proc. SP1E, Vol. 7557, 75570M, 2010 /. This source of information is the closest technical solution to the proposed invention.

The object of the invention is to increase the efficiency of multispectral imaging by increasing the number of images obtained from a single RGB color image data array corresponding to different spectral bands.

DETAILED DESCRIPTION OF THE INVENTION

To achieve this goal, a new method of processing the output data of digital RGB sensors and a device that implements it are proposed. The method differs from the nearest known technical solution in that:

(i) The discrimination level of RGB output signals is not fixed but variable. This allows for the addition of six known spectral intervals (not including the R, G, and B bands) to divide new intervals, thereby increasing the number of multispectral images obtained from single color image data;

(ii) For spectral selection, the intersections (BG, GR or BR) of the two spectral sensing curves at fixed wavelengths are additionally used. At these intersections, new spectral intervals are emitted which satisfy the conditions κBi-Gli <a (1) and / or jGi-R j J <b, (2) and / or | B _i -R _i | <c, (3) , b and c are controlled variable fixed numbers.

a brief description of the drawings

Fig. 1 shows typical relative spectral sensitivity curves for R, G, and B channels of an RGB sensor and illustrates a way in which variable signal discrimination level S can be used for additional multispectral image selection.

FIG. 2 shows how the intersections of the relative spectral sensitivity curves of the R, G, and B channels can be used for additional multispectral image selection.

FIG. Fig. 3 is a block diagram of a proposed multispectral imaging device ............

The essence of method (i) is illustrated in FIG. Relative spectral sensitivity curves for sensor R, G, and B channels, provided by the manufacturer or experimentally measured, are amplified so that the highest recorded signal value does not exceed the maximum channel output signal, such as 255 in an 8-bit system. such that the sensor operates in a linear mode, i.e., that the numerical values of Rj, Gj and Bi signals (for example, in the range 0 to 255) are proportional to the intensity of the optical signal being received. It then captures a single digital color image of the object, each of which contains pixels Rj, Gi and B in the data array; Signal values. These signal values are compared with the Sd value of the variable discrimination level and only those above the Sj level are used in further analysis.

Depending on the value of Sj, the following situations are possible: (a) signals are recorded in the i-pixel simultaneously in two color bands;

b) In an i-pixel, signals are recorded in only one color band, ie only the R, G or B channel. Situation (a) shows that the spectral range of the radiation received in the i-pixel is within the overlapping area of the spectral sensitivity curves of the two color bands. This is illustrated in Fig. 1. For example, the simultaneous registration of the values of Bj and Gj at the lowest discrimination level So means that the spectral range of the radiation incident on this pixel is between λ ₆ and λ ^ (the corresponding wavelengths of points 6 and 14, respectively). Depending on which value (Bj or Gj) is greater, narrower ranges of the perceived spectrum can be specified - either λ (, ... λιο if Bj> Gj, or λιο - λΐ4 if Bj <Gj). the simultaneous recording of Bj and Gj values at level S2 makes it possible to determine that the spectral range of the radiation incident on the i-th pixel is narrowed (between kg and λΐ2), or either λ ₉ ... λιο for Bj> Gj or λιο ··· λη if Bj <Gj. Similarly, the simultaneous recording of the values of Rj and G; at Si indicates a perceived spectral range λιγ ... λΐ8 for Ri <Gj, or λι§ ... λι ₉ for Rj> Gj , and so on.

Situation (b), in turn, points to one of two other possibilities. First, the recorded spectral range may correspond to one of the "peak" areas of the B, G, or R curves for the region λ ₄ ... λιο (for the B-channel at S3), for the region λιο - λιβ (for the G-channel at S3) or for the region λ ^. .. λιι (for R-channel at S2). By increasing the level of discrimination, the recorded spectral range narrows - for example, at level S4, it is only possible to register radiation in the λ5 ... Έ spectrum band and in the G-channel only in the λπ ... λΐ5 band. When the discrimination level reaches the peak of the B, G, or R band, the appearance of the signal in the respective band indicates the detection of monochromatic radiation (at the wavelength corresponding to the maximum of the band) in the given pixel. Second, signal registration in only one of the color bands at lower levels of discrimination suggests that the spectrum interval recorded in a pixel is not within the overlap of the two bands. For example, the discrimination level Si only Bjoslā detected signal corresponding to the spectrum range λ ₂ ... λ ", only the G-band spectrum to detect a signal corresponding to the interval λη ... λπ and only R-band signal detected - spectrum range λ ₁₉ ... λ ₂ 2 · /

Thus, the proposed solution ensures that the discrimination level S of the RGB output signals is not fixed but variable, which offers ample opportunities for selection of different spectral intervals for multispectral imaging.

The essence of method (ii) is illustrated in FIG. The three possible intersections of the RGB spectral sensitivity curves have fixed wavelength values. If there are pixels in the RGB image that have Gj = Bi, Rj = Gj, or Ri = B _b, then it must be concluded that the monochromatic radiation with the wavelength of the corresponding intersection is recorded in these pixels. In turn, it is proposed to create new spectral ranges around these three fixed wavelengths for multispectral imaging by imposing limits on the absolute value of the difference between the values of the signals recorded in the intersecting bands:

j Bj-Gi l <«and / or J Gj-Rj J <b and / or | Bj - Rj <<c.

Under these conditions, BG allocates the spectral band% a ... λ ^ GR around the intersection to the spectrum band λβ ... Af and BR to the intersection point the spectrum band Xc ... λ _ά . By changing the values of a, b, and c, the wavelengths of the spectral bands assigned to the multispectral display are also changed accordingly.

One embodiment of either of the above methods is illustrated in FIG. 3. It shall consist of:

- a multispectral light source (1), such as a set of several light emitting diodes and / or laser diodes emitting in different spectral bands, which irradiate the object being imaged (2), such as the skin surface;

- a digital RGB sensor (3) equipped with a lens, which converts the image of the object into a digital format, assigning each pixel a specific set of R, G and B values, and storing the complete RGB data array in the storage device (4);

- a converter (5) which converts the data of this array into spectral intensities together according to a selected level of signal discrimination imposed by the discriminator (6);

- an image selector (7) for performing a multispectral analysis using a suitable program, distributing images corresponding to each selected spectral interval from the original RGB data array;

- output devices (8), such as a computer monitor, to which the resulting set of multispectral images is output.

Further processing of multispectral images is possible by known methods.

Claims (5)

Claims 1. A method for obtaining a multispectral image using a digital RGB sensor from a monochrome image data array, illuminating an object simultaneously in multiple spectral bands, identifying and then comparing the values of the Rj, Gi, and Bj signals recorded at each pixel i of the image. with externally imposed signal discrimination level S, which at any wavelength sensor in the spectral sensitivity range only allows registration of one or two R, G and / or B bands, differing in that, in order to increase the number of multispectral images, S values are continuously resized to the highest possible values for the R, G, or B band signals, provided that the photosensitivity linearity of the sensor is maintained within this interval, and two situations are analyzed depending on the S value, where signals in two color bands (B and G, G and R or B and R) or when they are registered ie when a single color band is used, and the analysis identifies the range (s) of spectrum within which the radiation emitted by that pixel falls. A method for obtaining a multispectral image according to claim 1, characterized in that the spectral selection further uses the intersections of two spectral sensitivity curves (BG, GR or BR) at fixed wavelengths and / or extracts new spectral intervals around these intersections, using the conditions: B Bj - Gi <a, | Gj - Rj <b and / or J Bi - R j J <c, where a, b and c are controlled variable fixed numbers. The method for obtaining a multispectral image according to claim 1 or 2, characterized in that each of the identified spectral intervals or fixed wavelengths produces a single narrow-band image from pixels that corresponds to that spectral interval or fixed wavelength. length. The method of obtaining multispectral images according to claim 3, characterized in that narrowband spectral images are compared, divided, subtracted and / or otherwise mutually manipulated at the identified spectral intervals. A device for obtaining multispectral images according to the method as defined in any one of claims 1 to 4, comprising a multispectral light source, an objective digital RGB sensor, an RGB data array storage device, a converter that converts RGB data into a set of spectral intensities according to a selected signal discrimination level, an image selector that outputs images corresponding to each selected range of spectrum, and an output device such as a computer monitor.