Method for determining a texture of an object and associated devices for assisting with visualization of surface roughness

Abstract

The present invention relates to a method for determining a texture of an object (12), the method being implemented by a computer (16) of a system (10) for determining the texture of an object (12) and comprising a step of: - obtaining a set of images of the object (12), each image of the set being an image of the object (12) in a respective acquisition spectral band, each acquisition spectral band having at least one overlap with another acquisition spectral band, - selecting a reference image from the images of the obtained set, - subtracting the reference image from an image obtained from the one or more other images of the set of images, and - determining the texture image from the subtracted image.

Description

[0001]TITLE: Method for Determining the Texture of an Object and Associated Devices to Aid in Visualizing Surface Roughness. The present invention relates to a method for determining the texture of an object to aid in visualizing surface roughness. The present invention also relates to the devices involved in implementing such a determination method, namely a computer and a determination system. In the field of geology, it is desirable to be able to know the geological composition of the subsoil in order to deduce, in particular, the presence of oil or gas reservoirs, fluid pressure, and the presence of contaminants, among other things. For this purpose, it is known to analyze images taken from the surface to determine its texture. The texture of a rock defines the size, shape, and arrangement of the grains or crystals that compose it. This texture influences the surface roughness of the rock.Knowledge of texture allows, in particular, the identification of the presence of one or more specific minerals. Texture determination is generally performed by applying a statistical function to the image. Typically, an entropy filter or a standard deviation filter is used. However, all these functions are complex to implement and, therefore, not very compatible with use at the image acquisition site. There is thus a need for a method of determining the texture of an object from images that is easier to implement.To this end, the description concerns a method for determining the texture of an object. This method is implemented by a computer within a system for determining the texture of an object and comprises a step of: - obtaining a set of images of the object, each image in the set being an image of the object on a respective spectral acquisition band, each spectral acquisition band having at least one overlap with another spectral acquisition band; - selecting a reference image from among the images in the set obtained; - subtracting the reference image from an image obtained from the other image(s) in the image set; and - determining the texture image from the subtracted image. Such a method for determining the texture of an object thus makes it possible to determine a texture image and therefore corresponds to a method for determining a texture image of the object.Depending on other advantageous aspects, the determination method comprises one or more of the following features, taken individually or in all technically possible combinations: - the image set comprises at least three images, the method further comprising a step of combining images other than the reference image to obtain a combined image in the spectral band of the reference image, the combined image being the image subtracted from the reference image during the subtraction step. - each image corresponds to an observation channel of a camera used to obtain the image, the combination step comprising the application of a square matrix comprising 9 elements to images derived from the images corresponding to an observation channel - the matrix has one column in which the value of each element is equal to 1, the other elements being zero. - the determination step is a scaling of the subtracted image.- Scaling involves applying a different scaling function from one observation channel to another. - Scaling involves applying a scaling function common to all observation channels. - Frequency bands are chosen from a visible channel, a red channel, a blue channel, and a green channel of a camera. - Frequency bands are chosen from the ultraviolet and visible spectrum.The description also relates to a calculator for a system for determining the texture of an object, the calculator being able to: - obtain a set of images of the object, each image in the set being an image of the object on a respective spectral acquisition band, each spectral acquisition band having at least one overlap with another spectral acquisition band, - select a reference image from among the images in the set obtained, - subtract the reference image from an image obtained from the other image(s) in the set of images, and - determine the texture image from the subtracted image.The description also relates to a system for determining the texture of an object, the determination system comprising: - a camera suitable for acquiring a set of images of the object, each image in the set being an image of the object on a respective spectral acquisition band, each spectral acquisition band having at least one overlap with another spectral acquisition band, and - a computer as previously described.The invention will become clearer upon reading the following description, given solely by way of non-limiting example, and made with reference to the drawings in which: - Figure 1 is a schematic representation of an example of a system for determining the texture of an object, - Figure 2 is a flowchart of an example of the implementation of a method for determining the texture of an object, - Figure 3 is an example of a set of images that can be used as input for the method of determining Figure 2, - Figures 4 to 9 are images obtained at several stages of an example of the implementation of the method for determining Figure 2, and - Figures 10 to 14 are images obtained at several stages of another example of the implementation of the method for determining Figure 2. A system for determining the texture of an object 12 is schematically illustrated in Figure 1.The determination system 10 thus seeks to better visualize the surface of the object 12 in order to clearly observe variations in surface roughness. Any object exhibiting a perceptible variation in surface roughness can be considered here. The remainder of the text focuses primarily on a geological object such as a rock, but it is entirely conceivable to use this determination system 10 in other contexts, for example, to observe pieces of wood. The determination system 10 comprises a camera 14 and a computer 16. The camera 14 is capable of acquiring two or more images of the same object 12 in a fixed position of the camera 14 and across several spectral bands (multiple channels), regardless of lighting conditions. By way of non-limiting example, the spectral bands are visible and ultraviolet, or visible and the red, green, and blue channels of a visible camera 14.It is assumed here that camera 14 is a visible camera which therefore records the three channels during an acquisition of an image of object 12. We will note (I. R , I G , I B The three channels of an image I. Each of these channels can be represented by a matrix whose dimensions are those of the image in pixels, say m × n (m and n being two non-zero integers), so a representation of the entire image is a three-dimensional array, that is, an array of dimension m × n × 3. Colors can be encoded with varying degrees of precision. Image files generally encode each color using 8 or 16 bits (i.e., 24 bits per pixel or 48 bits per pixel), which means that each pixel of the image has a choice of 256 colors (2 8 ) or 65536 colors (2 16It is assumed that all the images considered in the following section have identical encoding, even if preprocessing is performed before implementing the procedure described below. For example, camera 14 operates in autofocus mode, which means that the focal length varies very slightly from one image capture to another. Alternatively, the focal length is maintained from one image capture to the next. This can be useful for visualizing depth, since the focal area can include elements slightly above or below the targeted object 12.The computer 16 is an electronic circuit designed to manipulate and / or transform data represented by electronic or physical quantities in registers of the computer 16 and / or memories into other similar data corresponding to physical data in register memories or other types of display, transmission, or storage devices. As specific examples, the computer 16 is implemented as a programmable logic component, such as an FPGA (Field Programmable Gate Array), or as an integrated circuit, such as an ASIC (Application Specific Integrated Circuit). Alternatively, when the process is implemented as one or more software programs, i.e., as a computer program, also called a computer program product, it is further capable of being recorded on a computer-readable, unrepresented medium.A computer-readable medium is, for example, a medium capable of storing electronic instructions and being connected to a computer system bus. Examples of such media include optical discs, magneto-optical discs, ROMs, RAM, any type of non-volatile memory (e.g., FLASH or NVRAM), or magnetic cards. A computer program containing software instructions is then stored on the readable medium. The operation of the computer 16 of the determination system 10 is now illustrated with reference to Figure 2, which shows an example of implementing a process for determining the texture of an object 12. The determination process comprises a retrieval step E20, a selection step E22, a combination step E24, a subtraction step E26, and a determination step E28. During the retrieval step E20, the computer 16 obtains a set of images.The images were previously acquired by camera 14. In the example described, each sensor acquires an image of object 12, preferably taken under the same shooting conditions. The computer 16 thus obtains images of the same object 12 taken across several spectral bands. Each spectral band overlaps with at least one other spectral band. At the end of the acquisition step E20, the computer 16 therefore has four images: a red image, a green image, a blue image, and a white image. An example of these images is visible in Figure 3, with the white image in the upper left, the red image in the upper right, the green image in the lower left, and the blue image in the lower right. During the selection step E22, the computer 16 selects a reference image, which in this case is the white image. More generally, the computer 16 selects as a reference image the image corresponding to the widest spectral band.The E24 combination step aims to create reconstructed images. The terms "reconstructed" or "recombined" are also used to refer to reconstructed images. In the example corresponding to Figure 4, the E24 combination step comprises a conversion substep and an E24 combination substep. During the conversion substep, the red, green, and blue images are converted into grayscale images. These latter images correspond to a representation of luminance. The computer 16 combines the three channels to produce three images, labeled GrR, GrG, and GrB, each with a size of mxn. The combination is performed by a linear combination of the channels according to the following formulas: Where: •^^, ^^^ and ^^ are coefficients between 0 and 1, • ^^, ^^^ and ^^ are coefficients between 0 and 1, and • ^^, ^^^ and ^^ are coefficients between 0 and 1. Furthermore, the notation ^ ^corresponds to the Z channel of the image taken under X-ray light. Typically, ^ ^^is the green channel of the image taken under red light. Preferably, the coefficients previously grouped into three groups are each equal. According to a particular embodiment, the following values ​​are chosen: •^^ = ^^^ = ^^ = 0.2989 or ^^ = ^^^ = ^^ = 0.299, • ^^ = ^^^ = ^^ = 0.587, and • ^^ = ^^^ = ^^ = 0.114. The different images GrR, GrG, and GrB are visible in Figure 4. During the combination substep, the calculator 16 combines the images GrR, GrG, and GrB thus obtained to obtain a new color image with a dimension of ^ × ^ × 3. This can be written mathematically as follows: ^:(^^,^^^,^^) = (^^^^^,^^^^^^,^^^^^) Where: •^ denotes the new color image, • ^^ is the first channel of the new color image, • ^^ is the second channel of the new color image ^ , and• ^^ is the third channel of the new color image. The new image thus obtained is the image in Figure 5.During the subtraction step E26, the calculator 16 subtracts the combined image from the reference image (the white image in this example). Such an operation is mathematically written as follows: ... Typically, for a set of values ​​between 0 and 255 (i.e., a modulo of 256), if ^. ^ = 10 and ^ ^ = 20, then ^ ^ = -10 which, with the modulo 256, becomes ^^= 246 (indeed, 255 corresponds to -1 here). The image thus obtained, corresponding to the difference vector, is visible in Figure 6. During the determination step E28, the calculator 16 determines the texture image from the subtracted image. A texture image here is an image that represents the surface roughness of object 12. To be more precise, it is an image where the focus surface is taken, and only small variations in altitude around this surface are considered. Indeed, for very large variations in roughness, these are out of focus, resulting in no visible details. According to the example described, the determination step E28 is a scaling of the texture image. The images obtained by this operation are visible in Figure 7. In the example described, the scaling corresponds to an adjustment of the dynamic range.Depending on the embodiments, scaling is implemented globally (top image of figure 7) or channel by channel (bottom image of figure 7). In each case, this allows us to obtain a scaled texture image where each component (on a respective channel) is denoted (^^,^^^,^^). Thus: ^^ = ^^.^^ + ^^^^^ = ^^^.^^^ + ^^^^^ = ^^.^^ + ^^ Where: •^^ and ^^ are two scaling coefficients for the red channel (respectively the first and second coefficients), •^^^ and ^^^ are the two scaling coefficients for the green channel, and • ^^ and ^^ are the two scaling coefficients for the blue channel. In the first case, corresponding to global scaling, each first scaling coefficient is the same from one channel to another, and similarly, each second scaling coefficient is the same from one channel to another.Calculator 16 then determines the scaling coefficients using the following formulas: Where: • the index i denotes the channel, i being R, G, or B in the example described, • ^^^(^) denotes the largest value of the matrix ^, and • ^^^(^) denotes the smallest value of the matrix ^. According to another embodiment, the correction applied depends on each channel. Thus, calculator 16 obtains the scaling coefficients using the following formulas: Where: • the index i denotes the channel, i being able to be R, G or B in the example described, • ^^^(^^) denotes the maximum value of the matrix corresponding to ^^ , and • denotes the minimum value of the matrix corresponding to ^^. The difference between the two techniques just described becomes clear when expressing the minima and maxima since: ^^^(^) = ^^^(^^^(^^), ^^^(^^^),^^^(^^)) and ^^^(^) = ^^^(^^^(^^), ^^^(^^^),^^^(^^)). To clearly highlight the advantage of the process, it is helpful to compare Figures 8 and 9. Figure 8 shows the image of object 12 seen in white light, while the two Figures 9 show the texture images obtained after implementing the process (top image: global scaling technique, and bottom image: channel-by-channel scaling technique). In Figure 8, black rocks in visible light are represented by dots, red rocks by crosses, and white rocks by nothing.The images in Figure 9 clearly distinguish these three types of rocks: black rocks appear flat with abrupt edges, red rocks appear flat but with smoother edges, and white rocks appear with relief and smooth edges. The method thus effectively differentiates the nature of the rocks by texture. Furthermore, the method is easy to implement, as the associated computational load is low. This method could be advantageously used to differentiate between two rocks of the same color but different compositions, and thus annotate such images to serve as a training dataset for an artificial intelligence algorithm used for edge recognition or automatic rock identification. Other implementations of the identification method offering the same advantages are also conceivable.A first example is described with reference to Figures 10 to 14. In this case, the E24 combination step is implemented differently. The E24 combination step comprises an extraction substep and an E24 combination substep. The extraction substep consists of extracting each channel corresponding to an object 12 illuminated by the corresponding spectral band. This means that a red image... ^ ^ is extracted from the red channel of the red image, which is a green image ^^ ^^ is extracted from the green channel of the green image and a blue image ^ ^ is extracted from the blue channel of the blue image. Formally, this amounts to the conversion sub-operation in which the coefficients are fixed as follows: •^^ = ^^^ = ^^ = 1 • ^^^ = ^^ = ^^ = ^^ = ^^ = ^^^ = 0. The different images^ ^ , ^^ ^^ and ^ ^are visible in Figure 10. The combination substep is similar to the previous substep, the new mathematical expression becoming: ^:(^^,^^^,^^) = (^^,^^^^,^^). The new image thus obtained according to the second example is the image in Figure 11. The image obtained after implementing the subtraction step E26 is shown in Figure 12, while the images obtained after implementing the determination step E28 are shown in Figure 13 (top figure, case of correction on all channels, and bottom figure, case of correction channel by channel). Figures 14 are equivalent to Figures 9 and therefore correspond to the texture images obtained at the end of this determination process. These images should therefore be compared to the image in Figure 8. Here again, it is clear that the texture images allow us to identify the differences in texture between the different rocks.It is also possible to improve certain steps of the process by adding additional operations. For example, according to a more elaborate embodiment, given the values ​​of the components. ^ ^ , ^ ^^ , ^ ^ , ^ ^ , ^ ^^ And ^ ^Since previous values ​​can be negative, the E26 subtraction step of the process includes a preliminary operation to convert each component into a format that allows each value to be expressed as a signed integer. In another example, an additional multiplicative coefficient can be applied to each of the scaling coefficients. Such a multiplicative coefficient is chosen, for example, to ensure that the scaling coefficients have values ​​spanning a larger range. This example is compatible with both scaling techniques (global or channel-by-channel).

Claims

CLAIMS 1. A method for determining the texture of an object (12), the method being implemented by a computer (16) of a system for determining (10) the texture of an object (12) and comprising a step of: - obtaining a set of images of the object (12), each image of the set being an image of the object (12) on a respective spectral acquisition band, each spectral acquisition band having at least one overlap with another spectral acquisition band, - selecting a reference image from among the images of the set obtained, - subtracting the reference image from an image obtained from the other image(s) of the image set, and - determining the texture image from the subtracted image. 2.A method of determination according to claim 1, wherein the image set comprises at least three images, the method further comprising a step of combining images other than the reference image to obtain a combined image in the spectral band of the reference image, the combined image being the image that is subtracted from the reference image during the subtraction step.

3. A method of determination according to claim 2, wherein each image corresponds to an observation channel of a camera (14) used to obtain the image, the combination step comprising applying a square matrix comprising 9 elements to images derived from the images corresponding to an observation channel.

4. A method of determination according to claim 3, wherein the matrix comprises a column in which each element has a value of 1, the other elements being zero. 5.A method for determining the value of an image according to any one of claims 1 to 4, wherein the determination step is a scaling of the subtracted image. A method for determining the value of an image according to claims 3 and 5, wherein the scaling involves applying a different scaling function from one observation channel to another.

7. A method for determining the values ​​according to claim 5, wherein the scaling involves applying a scaling function common to all observation channels.

8. A method for determining the values ​​according to any one of claims 1 to 7, wherein the frequency bands are selected from a visible channel, a red channel, a blue channel, and a green channel of a camera (14).

9. A method for determining the values ​​according to any one of claims 1 to 7, wherein the frequency bands are selected from the ultraviolet and visible spectrum. 10.Calculator (16) of a system for determining (10) the texture of an object (12), the calculator (16) being suitable for: - obtaining a set of images of the object (12), each image in the set being an image of the object (12) on a respective acquisition spectral band, each acquisition spectral band having at least one overlap with another acquisition spectral band, - selecting a reference image from among the images in the set obtained, - subtracting the reference image from an image obtained from the other image(s) in the image set, and - determining the texture image from the subtracted image. 11.A system for determining the texture of an object (12) (10), the determination system (10) comprising: - a camera (14) suitable for acquiring a set of images of the object (12), each image in the set being an image of the object (12) on a respective spectral acquisition band, each spectral acquisition band having at least one overlap with another spectral acquisition band, and - a computer (16) according to claim 10.

Images