LV14782B - Method for imaging of spectral attenuation at a number of wavelength bands - Google Patents

Abstract

The invention relates to spectral imaging, in particular to imaging of spectral reflectance distribution at several fixed wavelengths using a single RGB image data set. In the proposed method for obtaining several spectral reflectance images, they are extracted from a digital RGB image data set at the wavelengths corresponding to those used at polychromatic illumination of the target, providing linearity of the photo-response, using the reference reflector and taking into account the overlapping of the RGB photo-detector spectral sensitivity bands, where the spectral reflectance is calculated for each image pixel or group of pixels using a calibrated grey reference reflector with reflection coefficients at all exploited wavelengths less than those of the white reference and by value is close to the largest reflection coefficient of the displayed object area. In case of different intensities of the illumination spectral lines, corrections for spectral reflectance values are proposed.

Description

Description of the Invention

Technical field

The invention relates to spectral imaging, in particular to the representation of spectral attenuation coefficients of a surface reflection at multiple fixed wavelengths using data from a single digital RGB image.

Known state of the art analysis

The spectral attenuation or spectral reflectance of reflection (WO 2013135311 Al, 2013) k (X) is the ratio Ι (λ) / Ι ₀ (λ), where Ι (λ) is the intensity reflected from the object's surface area at one specified wavelength λ and Ι ₀ (λ) is the intensity of reflection from a fully reflective reflector (so-called white reference) under the same lighting conditions at wavelength λ.

Digital sensors, such as two-dimensional photometric matrices, are often used to produce spectral images, complete with filters that pass through several spectral regions, such as a rotating disk with a set of different color filters (EC Ruvolo et al., Proc. SPIE, Vol. 7548, 75480A, 2010) uses acoustic or liquid crystal filters spectrally read by electrical signals (http://www.usgs.gov/science/science.php?term=765). Digital RGB sensors incorporate three-color (blue, green, and red) spectral filters, which in combination with external spectral filters provide a greater number of spectral images (US 7612822 B2, US 2009290124 (Al), JP 2008136251 (A)). Another way to obtain a spectral array of images is to use multiple light sources emitting in different spectral bands (e.g., LEDs of different colors - WO 2008093988 (Al)) to sequentially illuminate the subject by capturing its own image in each illumination spectrum band.

There are known methods for separating multiple spectral images from a single digital RGB image data array. For example, for each pixel or pixel of an image, the numerical values Ri, Gi, and Bi of the signals recorded in the red (R), green (G), and blue (B) spectral bands are determined and used to produce spectral images in the red, green and blue and / or perform multispectral analysis by dividing these sub-images by subtracting one from the other, 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 values of R, G, B signals at any wavelength in the sensor spectral sensitivity range, it is possible to logically analyze six more narrow spectral intervals in the RGB data array (J. Spīgulis et al., Proc. SP1E, Vol. 7557,75570M, 2010). By changing the discrimination level values, the number of spectral intervals to be distributed from one digital RGB data array increases significantly (W0 2012/002787 Al, 2012).

These and other technical solutions can be accomplished by polychromatic illumination of the subject at the time of capturing a single RGB image, i.e. illuminating it simultaneously in as many spectral bands as possible. A special advantage is illumination, which has only narrow spectral lines with fixed wavelengths in the spectrum, which allows for more accurate restoration of the spectral images, as compared, for example, to measurements in multi-band LED illumination. The ability to map the distribution of relative concentrations of skin oxyhemoglobin in bichromatic laser illumination using single RGB imaging data has been demonstrated (J. Spigulis et al., Proc. SPIE, Vol. 7557, 75570M, 2010). However, the method does not allow the determination of the values of the spectral attenuation coefficients for a given reflection at each pixel at both fixed wavelengths.

Since the three spectral sensitivity curves of RGB image sensors are usually overlapping, none of the color channel signals adequately describes the intensity of the received monochromatic radiation. The effect of inter-channel crosstalkF at a fixed wavelength can be corrected if the spectral sensitivity curves of the RGB channels are known and a linear photo capture mode is provided (J.Spigulis et al., Proc. SPIE, v. 8216, 82160L , 2012).

Patent LV 14532 B (2012) provides a method and device for obtaining multiple spectral images from a single RGB data array according to the number of spectral lines of illumination. This solution allows the determination of k (z) numerical values at each fixed image wavelength (pixel) at a number of fixed wavelengths. displaying k (X) values at each of the selected wavelengths with a "crosstalk effect correction". This is achieved by inserting a fully reflective element (white reference) in the area of the object being displayed, whose reflected signals are compared with the reflected signals of different areas of the object. This solution, which is the closest analogue to the invention, also has disadvantages:

(1) The linearity limit of photo-capture is determined by the reflected signal from the white reference, which in real situations can be significantly more intense than the reflected signals from the surface of the subject (eg human skin). Therefore, the signal-to-noise (S / N) ratio and the accuracy of the detectable k (Zi) values, where i is the number of illumination spectral lines used, are relatively low in the pixels of the subject image. For example, if the detector detects 256 intensity levels per pixel, the average noise level is 20 and the white reference area has a reflected signal level of 250, and the object surface reflected levels are between 40 ... 80, then the S / N value for the reference signals is 15 ( relative error of 8%) but for object signals only

2 ... .4 (relative error 25 ... 50%), which represents the relative error of the detectable k (Li) in the sum

Within the range of 33 to .58%;

(2) Corrections for interchannel interaction (.crosstalk) at fixed wavelengths are calculated assuming the same intensities for all illumination spectral lines. The actual light sources available may also vary, thus reducing the accuracy of the detectable k (Xi) values. Using smoothing filters in such situations is quite complicated and also an expensive solution.

Object, nature and detail of the invention

The object of the present invention is to increase the quality of representation of the spectral attenuation k (X) and the accuracy of determination of k (L) values using a single RGB image data array for a polychromatic line under spectral illumination.

To increase the accuracy of the reflection of the surface reflections of the object, it is proposed to replace the white reference with a calibrated gray reference that reflects all wavelengths of illumination spectral lines used and has a reflection coefficient close to the reflection level of the object. signal values. The gray reference reflectance coefficient k '= I / o, where I is the reflectance intensity from the gray reference and Io is the reflectance intensity from the white reference, can be calibrated at fixed wavelengths to an accuracy of less than 1%. The formulas of patent LV 14532 B, in which white reference signals are replaced by gray reference signals, are used to calculate the reflectance spectral attenuation values and the corrected spectral attenuation values are obtained by multiplying the values of ki, k ₂ , ..., k _n found. The positive effect of reducing the relative error of the reflected intensity measurements is illustrated by a numerical example. Assuming that the level of the reflection signal from the gray reference is 250, the levels of reflection from the subject are in the range 150 ... 225 and the noise level of the image sensor is 20. In this case, the respective S / N values for referencei and object will be 12.5 and 7.5 ... 11.3 , which corresponds to relative errors of 8% and 9 ... 13%, respectively, but the relative error of the detectable k (Xi) values is only 18 ... 23%, ie more than 2 times lower than using the white reference (see above). example). The use of gray references also allows for a significant increase in the illumination level of the subject, maintaining the linearity of the photo acquisition and consequently improving the quality (contrast, resolution) of the spectral images to be obtained. In summary, replacement of the white reference with gray significantly improves the accuracy of the detectable k (Xi) values and the quality of the spectral attenuation images of the reflection.

The spectral lines intensities of polychromatic illumination may vary with each other during reflection spectral attenuation. In such cases, adjustments to the image processing calculation formulas are required because the signals received by all pixels of the image depend not only on the spectral sensitivity Sx of the image sensor at the selected wavelengths, but also on the relative intensities of the respective spectral lines of illumination.

In the case of bichromatic illumination, it is proposed to introduce a correction factor C12 = I1 / I2, which is the ratio of the luminous intensities at wavelengths λι and λ2, to multiply the relative photographic sensitivity Sr, Sg and / or Sb as defined in patent LV 14532 B. By multiplying ci2 * Sr, ci2 * Sg and / or ci2 * Sb by the respective values of Sr, Sg and / or Sb in the formulas for the calculation of the reflection spectral attenuation, the difference in spectral line intensities is observed in bichromatic illumination situations.

In trichromatic illumination, the correction of the intensity differences is similar: it is proposed to introduce correction factors C12 = I1 / I2, C13 = I1 / I3 and C23 = I2 / I3, which are the ratios of the luminous intensities at wavelengths λι,% 2 and% 3. The next step, with correction factors, multiplies the image sensor's relative photosensitivity at the appropriate wavelengths across all three photo-acquisition bands and replaces the relative photosensitivity used in patent LV 14532 B

S (Rf

S (R 3) ' _z S (R ₂ ) _{ς r} 5 ((7,) S (R ₃ )' ^0.2 S (G ₂ ) _ S (G f>

S (G ₃ ) ₌ S (G ₂ ) S (G ₃ ) 'B12

S (B ₂ ) ' ⁵¹³ S (B ₃ ) _S (B ₂ )

S (B ₃ ). Specifically, the adjusted values to be replaced are used

The values of Sri2, Sgi2, Sbi2, Sri3, Sgi3, Sbi3, Sr23, Sg23 and Sb23 in the known solution are as follows: ci2 * Sri2, ci2 * Sgi2, ci2 * Sbi2, ci3 * Sri3, ci3 * Sgi3, ci3 * Sbi3, C23 * Sr23, C23 * Sg23 and C23 * Sb23. In general, by dividing the images of the spectral attenuation of the n reflections from a single array of RGB data by spectral line illumination of n different intensities, the correction is to be made in the system matrix defined by LV 14532 B.

sensitivity at wavelengths λι, λ ₂ , ..., λ _η : Α =

G _n . In this situation, applying the intensities of the two spectral lines B ' _t to the correction is not rational, since such correction coefficients can be very numerous. It is proposed to first determine the relative intensities of each illumination spectral line relative to the most intense of them, ie positive values Ii ', I2',, In with values <1. Next, h 'multiplies all members of column 1 of the matrix, column members, etc .; multiplies all members of the matrix ncol by /. The A-matrix thus corrected can be used to determine the refined spectral attenuation values for each pixel or group of pixels in an object image with unlimited illumination spectral lines n.

Claims (4)

Claims 1. A method for representing reflective spectral attenuation at multiple wavelengths, wherein the spectral attenuation images are distributed from a single digital RGB image data array at wavelengths corresponding to the wavelengths of a polychromatic line spectrum illumination spectral line using a reference reflector and taking into account The overlap of bands, which differs by using a calibrated gray reference reflector for calculating the spectral attenuation of reflection in each pixel or group of pixels, which has a reflection coefficient at all illumination wavelengths lower than the white reference and is close to the highest reflection coefficient in the area of the object. The method of claim 1, further correcting errors in the spectral attenuation values of the reflection due to differences in spectral line intensities of the bichromatic illumination by introducing a correction factor ci ₂ = Ii / I ₂ , which is the ratio of illuminances to wavelengths λι and λ ₂ , and in the known formulas for the calculation of reflection spectral attenuation, the relative photosensitivity values of Sr, Sg, and / or Sb at the selected wavelengths are replaced by the product ci ₂ * Sr, ci ₂ * Sg and / or ci ₂ * Sb, respectively. The method of claim 1, further correcting errors in the spectral attenuation values of the reflection due to differences in spectral line intensities of the trichromatic illumination by introducing correction factors ci ₂ = Ii / I ₂ , C 13 = I ₁ / I ₃ and c ₂₃ = I ₂ / I ₃ , which are the ratios of luminous intensities at wavelengths λι, λ ₂ and λ3 to multiply the relative photosensitivity of the image sensor at the respective wavelength pairs and to replace these values by the known formulas for reflection spectral attenuation. 4. The method of claim 1, further adjusted, reflection spectral attenuation value errors polychromatic illumination n spectral intensities difference the first setting each light spektrāllīnijas relative intensity of the most intense of which, ie positive values h ', h', I _n 'with values less than or equal to 1 and then in the system matrix ^ 1 ^G 1 ^G 2 R 'formed by the spectral sensitivities of the R, G and B channels at wavelengths - ^ 2 ·· nj λι, λ ₂ , ..., λ _η , multiplied by Ii 'for all column 1 members, by I2' multiplied by all column 2 members, etc., incl. multiplying by I _n 'all members of column n of the matrix where n> 2.