[0001] The present invention relates to providing an illumination effect value of a volumetric dataset, and to a method of direct volume rendering of a volumetric dataset.
[0002] Visualization of a volumetric dataset may comprise a process which may be referred to as volume rendering. Volume rendering may, for example, be used as part of a medical imaging process, for example to visualize a volumetric dataset gathered by a data gathering process. In the field of medicine, volume rendering may allow for a radiologist, a surgeon or a therapist to understand and interpret volumetric data and may support communication with the patient and may also support education of medical practitioners.
[0003] Visualization of medical datasets may be used for example, for diagnosis, teaching, patient communication etc. Visualization of volumetric data is also applied in numerous other technical fields, for example in geological, industrial quality assurance and scientific simulations.
[0004] According to a first aspect of the present invention, there is provided a method of determining an illumination effect value of a volumetric dataset, the method comprising: determining, based on the volumetric dataset, one or more parameter values relating to one or more properties of the volumetric data set at a sample point; providing the one or more parameter values as inputs to an anisotropic illumination model and thereby determining an illumination effect value relating to an illumination effect at the sample point, the illumination effect value defining a relationship between an amount of incoming light and an amount of outgoing light at the sample point.
[0005] The anisotropic illumination model may be an ellipse-based model comprising one or more parameters defining a mathematical model of an ellipse, and the one or more parameter values relating to one or more properties at the sample point may be provided as input values for the one or more parameters defining an ellipse.
[0006] The one or more parameter values relating to one or more properties at the sample point may be values relating to a curvature at the sample point.
[0007] At least one of the values relating to the curvature at the sample point may be a principal curvature value or a principal curvature direction value.
[0008] The determined illumination effect value may be a specular term of a bi-directional reflection distribution function (BRDF).
[0009] The anisotropic illumination model may be defined according to the expression:
wherein:
f_{s,1} is the illumination effect value and is the specular term of the BRDF;
e_{T} and e_{B} are the parameters which define an ellipse;
N is a normal vector at the sample point;
L is a light vector at the sample point;
V is a viewing vector at the sample point;
H is a halfway vector being halfway between L and V and is defined as
F_{r} is a Fresnel term; and
φ_{h} is an azimuthal angle between H and a tangent vector T.
[0010] The anisotropic illumination model may be defined according to the expression:
wherein:
f_{s,2} is the illumination effect value and is the specular term of the BRDF;
e_{T} and e_{B} are the parameters which define an ellipse;
θ_{i} is an elevation angle between a normal vector N at the sample point and a light vector L at the sample point;
θ_{o} is an elevation angle between the normal vector N and a viewing vector V;
θ_{h} is an elevation angle between the normal vector N and a halfway vector H, wherein H is a vector halfway between L and V and is defined as
and
φ_{h} is an azimuthal angle between the halfway vector H and a tangent vector T.
[0011] The anisotropic illumination model may be defined according to the expression:
wherein:
f_{s,3} is the illumination effect value being determined and is the specular term of the BRDF;
e_{T} and e_{B} are the parameters which define an ellipse;
a is a constant defining a shininess of a material associated with the sample point; and
φ_{h} is an azimuthal angle between a halfway vector H and a tangent vector T, wherein H is a vector halfway between L and V and is defined as
[0012] Providing the one or more parameter values as inputs to the anisotropic illumination model may comprise providing one of the one or more parameter values as input values for one of the parameters
e_{T} and
e_{B}.
[0013] Providing one of the one or more parameter values as input values for each of
e_{T} and
e_{B} may comprise providing parameter values for
e_{T} and
e_{B} as follows:
and
or
and
or
and n,
wherein
and n, or
and n,
wherein
and
wherein
n is a normalisation value dependent on the anisotropic illumination model being used.
[0014] The method according to the first aspect may comprise providing principal curvature directions as inputs for the tangent vector
T and the bitangent vector
B, such that the tangent vector
T and the bitangent vector
B are each aligned with one of the principal curvature directions at the sample point.
[0015] According to a second aspect of the present invention there is provided a method of performing direct volume rendering of a volumetric dataset comprising: simulating a plurality of rays converging at a viewpoint and passing through the volumetric dataset; selecting a plurality of sample points in the volumetric dataset along each of the plurality of rays; classifying one or more optical properties of each of the sample points; determining, for a given sample point of the plurality of sample points, an illumination effect value relating to an illumination effect at the given sample point by the method according to the first aspect of the present invention; determining, for the given sample point, a value of one or the optical properties at the sample point based on the classification of the optical property at the sample point and the determined illumination effect value at the sample point; and accumulating along each simulated ray values of the optical properties determined for each sample point along the ray to thereby obtain an accumulated value of the optical properties for use in rendering the volumetric dataset.
[0016] According to a third aspect of the present invention there is provided a set of machine-readable instructions which when executed by a processor cause a method according to the first aspect or the second aspect to be performed.
[0017] According to a fourth aspect of the present invention there is provided a machine-readable medium comprising a set of machine-readable instructions according to the third aspect.
[0018] According to a fifth aspect of the present invention there is provided apparatus comprising a processor and a storage comprising a set of machine-readable instructions which when executed by the processor cause the processor to perform a method according to the first aspect or the second aspect.
Brief Description of the Drawings
[0019]
Figure 1 illustrates a flow chart representation of an example method of determining an illumination effect value for a volumetric dataset;
Figure 2 illustrates schematically part of an example method of rendering a volumetric dataset, according to an example;
Figure 3 illustrates a flow chart representation of a method of rendering a volumetric dataset, according to an example;
Figure 4a and Figure 4b illustrate schematically a portion of a volumetric dataset and associated vectors for use in an example method of determining an illumination effect value; and
Figure 5 illustrates schematically a system comprising an apparatus for rendering a volumetric dataset, according to an example.
[0020] Figure 1 illustrates a flow chart representation of an example method 100 of determining an illumination effect value of a volumetric dataset. The method 100 comprises, at block 101, determining one or more parameter values relating to one or more properties of the dataset at a sample point. Examples of parameter values relating to properties of the dataset and which may be determined at block 101 include values relating to curvature at the sample point, but also may relate to other values relating to the sample point, as will be discussed below.
[0021] At block 102, the method comprises providing the one or more parameter values relating to one or more properties of the dataset at the sample point as inputs to an anisotropic illumination model. As referred to in examples herein, an illumination model is a mathematical model which may be used to simulate the interaction of light with a point in a volume. An illumination model may be defined by a mathematical expression for providing an illumination effect value at the sample point. In some examples, a mathematical expression defining an illumination model comprising one or more parameters, for which parameter values relating to the sample point may be input such that the expression provides an illumination effect value for the sample point. Types of illumination models include isotropic and anisotropic models, and examples of anisotropic illumination models which may be used in example methods such as the method 100 will be discussed below.
[0022] By inputting the one or more parameters relating to the sample point into the anisotropic illumination model, the method 100 determines an illumination effect value for the volumetric dataset. The illumination effect value defines a relationship between an amount of incoming light and an amount of outgoing light at the sample point. For example, the illumination effect value may in one example be a value at least partly defining a bi-directional reflection distribution function at the sample point. Examples of illumination effect values which may be determined by the example method 100 will be discussed below.
[0023] The example method 100 may be performed, for example, during a method of rendering, for example direct volume rendering, of a volumetric dataset.
[0024] Figure 2 illustrates schematically a volumetric dataset 200 and an example method of rendering the volumetric dataset 200. The volumetric dataset 200 may be referred to herein as the volume 200 or the dataset 200.
In one example the volumetric dataset 200 comprises a discrete sampling of a scalar field. For example, the volumetric dataset 200 may comprise a 3D medical dataset. Such a medical dataset may be received by loading from a memory, sensors, and/or other sources. Such a medical dataset may represent a part of a patient, for example a human or animal patient. In general, any scanning modality which will produce a 3D volumetric dataset may be used to produce the volumetric dataset 200. For example, the scanning modality may comprise the use of computed tomography (CT), or of magnetic resonance imaging (MRI). In some examples a scanning modality comprising the use of positron emission tomography (PET), single photon emission computed tomography (SPECT), ultrasound, or another scan modality may be used. Scan data may be provided in the form of multiple two-dimensional (2D) scans or may be formatted from a 3D scan. In some examples, the volumetric dataset 200 is a DICOM dataset created by scanning at least a portion of a patient using a scanning modality. In other examples, values making up the volumetric dataset 200 may represent geological data (e.g. gathered using seismic data), or as part of industrial quality assurance (e.g. gathered using industrial x-ray scans). In other examples, the volumetric dataset 200 may comprise values representing an object produced via a scientific model rather than measured values representing a physical object.
[0025] In examples, the volumetric dataset 200 may comprise data formatted as a plurality of voxels 201. The voxels 201 may, for example, be in a uniform or non-uniform 3D grid, or may be arranged in some other type of geometry (e.g., polar coordinate format). Each voxel 201 may represent a scalar value, such as scalar value obtained by sampling a scalar field, as described above. The type of scalar value represented by each voxel 201 may be dependent on the means by which the volumetric dataset 200 is obtained. For example, where a CT scanner is used to produce the volumetric dataset 200, the dataset may comprise Hounsfield values. In this example, the volumetric dataset 200 comprises a representation of an object 210, which may be a representation of a portion of a medical patient or the like.
[0026] An example method of visualising the volumetric dataset 200 shown in Figure 2 comprises defining a viewpoint 10 with respect to the volumetric dataset 200. A viewing plane 20 is also defined and placed in front of the viewpoint 10. The viewing plane 20 comprises a number of pixels (not shown in the figures), e.g. arranged in a grid, and allows construction of a 2D visualisation of the 3D volumetric dataset 200, as viewed from the viewpoint 10. Some example methods of visualising the volumetric dataset 200 may be referred to as direct volume rendering. In an example, direct volume rendering comprises traversing a number of simulated rays 30, 31 through the volumetric dataset 200, wherein each of the rays 30, 31 intersects the viewing plane 20 and the rays 30 are convergent at the viewpoint 10. For example, one ray, or more than one ray, may be traversed through the volume 200 for each pixel of the viewing plane 20.
[0027] Each ray, e.g. ray 30, which is traversed through the volume 200 may allow determination of a value or set of values for display by a pixel of viewing plane 20 which is intersected by that ray. For example, a rendering algorithm may be employed which determines a value for display by the pixel via a calculation taking into account the path of the ray 30 through the volume 200. In some examples, a colour and brightness for the pixel may be determined via traversing a ray 30 through the volume 200 from the viewpoint 10 and calculating the effect on the ray 30 of accumulated optical properties of parts of the volume 200 which lie along a path of the ray 30. Such an example method may be referred to as direct volume rendering by ray casting.
[0028] In examples, points in the volume 200 are classified and assigned one or more optical parameters which define how that point of the volume 200 affects a ray 30 intercepting that point. For example, the effect of a voxel 201 of the volume 200 on a ray 30 may be defined by a set of parameters assigned to the voxel 201 and assigning optical properties to the voxel 201. In one example, a set of optical parameters assigned to each voxel 201 comprises an opacity and a colour. In examples, the set of parameters defining each voxel 201 is determined via use of a transfer function. A transfer function may assign optical parameters to the voxel 201 based on, for example, at least the scalar value of that voxel 201 of the volume 200. In some examples, additional properties related to the voxel 201, such as a gradient of the scalar values of the volumetric dataset 201 at the voxel, may be used as an input into a transfer function and therefore may affect the optical property/properties assigned to the voxel 201.
[0029] In other examples, a transfer function may assign to a given point in the volume 200 one or more of: a scattering coefficient, a specular coefficient, a diffuse coefficient, a scattering distribution function, a bidirectional transmittance distribution function, a bidirectional reflectance distribution function, and colour information. These parameters may be used to derive a transparency, reflectivity, surface roughness, and/or other properties of the surface of the given point. These surface material properties may be derived based on scalar values of the volumetric dataset at the rendering location, and/or based on user-specified parameters.
[0030] In examples, a plurality of sample points within the volumetric dataset 200 along the path of the ray 30 are selected. For example, sample points may be selected at regular intervals along the ray 30. In Figure 2, a first sample point 211 and a second sample point 212 along the ray 30 is shown, while the ray 31 comprises a plurality of sample points 213, 214. In examples, some sample points may lie outside of the volume 200 and thus will not contribute to the value/s for display by the pixel with which the ray is associated. It will be appreciated that in some examples sample point locations may be chosen using an algorithm which improves efficiency of the volume rendering method.
[0031] It should be noted that at least some of the sample points may not be coincident with a voxel 201 and as such calculations relating to a particular sample point, such as use of the transfer function, may employ interpolation to determine a scalar value at that sample point. For example, the first point 211 may not be coincident with one of the voxels 201. Trilinear interpolation, or another example method of interpolation, based on the scalar values of a set of voxels neighbouring the first point 211 may then be performed to determine an interpolated scalar value for the first point 211. Classifying of the point 211 may then comprise applying the transfer function to the interpolated value of the volumetric dataset 200 at the sample point 211. Similarly, a gradient of the scalar values of the volume 200 at the point 211 may be interpolated from neighbouring gradient values or may use interpolated scalar values and determine a gradient from those interpolated values. It will be appreciated that interpolation may be done before applying a transfer function or the like, or values resulting from application of such a function may themselves be interpolated for use in methods herein.
[0032] In this example and in other examples, a rendering algorithm, in determining an accumulated optical property, such as opacity and colour along the path of each ray 30, 31, models an effect on the values for accumulating for each sample point 211 of illumination of the volume 200. In this example method, the volume 200 is illuminated by a model light source 50.
[0033] In examples, the light source 50 may be a point source, a directional light source, or may comprise a light map. The simulation light source may also be any other kind of light source - e.g. a model of any object which emits light - or a combination of multiple different light sources. In some examples, parts of the volumetric dataset itself may emit light. In some examples, the light source may comprise a high definition light map. The light map in some examples may have six sides corresponding to outer sides of the volumetric dataset where the volume is cuboidal, for example.
[0034] Now with reference to Figure 3 an example method of rendering the volumetric dataset 200 is illustrated with a flowchart representation, starting at 301. As described with reference to Figure 2, the method shown in Figure 3 comprises traversing a plurality of rays 30, 31, originating at the viewpoint 10 through the volume 200. The rays 30, 31 have positioned along them a plurality of sample points, e.g. sample points 211, 212, 213, 214, etc. Each iteration of blocks 302 to 307 of the illustrated method, represents actions performed at a given sample point along one of the rays 30, 31 being traversed.
[0035] At block 302 it is determined whether the current sample point being considered is outside of the volume 200, i.e. at a point where the ray 30 upon which the sample point lies has already passed through the volume. If the answer to this is "yes", then the method ends with action 308, since this indicates that traversing of the ray 30 through the volume 200 is complete. If the current position at along the ray 30 is within the volume 200 then the volume may be sampled at block 303.
[0036] At block 304, in this example, the sample point being sampled is classified. This may comprise application of the transfer function at the sample point, as described above with reference to Figure 2, for example to determine optical properties, e.g. opacity and colour, of the sample point of accumulating along the ray 30.
[0037] At block 305 an illumination effect is applied at the sample point. Applying an illumination effect at the sample point, comprises determining an illumination effect value by performing a method according to example methods described with reference to Figure 2. The illumination effect value may then be used to modulate one or more of the one or more optical properties, such as colour and opacity, determined at block 304 by classification of the sample point. As such, a value of the one or more optical properties may be determined for the sample point based on the classification of the sample point and the determined illumination effect value. For example, a colour value determined for the sample point by classification of the sample point may be modulated according to the determined illumination effect value to produce a modulated colour value for the sample point. Further details of example methods of determining an illumination effect value at a sample point will be discussed in more detail below with reference to Figures 4a and 4b.
[0038] At block 306 the one or more optical property values determined at block 305 are accumulated into a sum of the one or more optical property values along the ray 30. At block 307 the method advances to the next sample point, e.g. advances from first sample point 211 to second sample point 212, and repeats block 302 (and blocks 303 to 307, where appropriate) of the above-described example method. The method thereby results in an accumulated value of the one or more optical properties determined for each sample point, to thereby obtain an accumulated value of the one or more optical properties for use in rendering the volumetric dataset. For example, as described above, the method described with reference to Figure 3 may allow accumulated values of opacity and colour along the ray 30 to be obtained which may be used for determining values for display by a pixel in the viewing plane 20.
[0039] Example methods of applying an illumination effect during example methods of volume rendering may add realism to an image rendered of the volume 200. For example, illumination effects may assist a user in interpreting objects, surfaces, and/or material boundaries represented by the volumetric dataset 200. For example, illumination effects may help to convey visual shape information of a rendered object to a user.
[0040] In examples, a local illumination model may be used to model the interaction of light with the volume 200. In general, when considering a particular point in the volume, such as the first sample point 211, a local illumination model may consider only light directly incident from the light source 50 on that particular point. A global illumination model, in contrast, may consider light which has arrived at the point in the volume from other points in the volume, for example via scattering or refraction. Local illumination models may, in some examples, involve a lower number of calculations for a given volume than a global illumination model, and therefore a local illumination model may provide less of a burden on computational resources than a global illumination model. Details of particular local illumination models will be returned to below.
[0041] In example methods of direct volume rendering, an overall contribution of colour from the point 211, taking into account the colour of the point 211 determined from the transfer function and any light scattered into or out of the ray 30 by the point 211, can be determined. This technique is derived from the physical approximation of the interaction of light with the volume which may simply be referred to as the rendering equation. In an example rendering algorithm, the overall effect of this point 211 on the ray 30 may be accumulated, e.g. via a summation operation, in order to solve the rendering equation. The algorithm may then comprise continuing to traverse the ray 30, moving to another sample point 212. The algorithm may thus integrate along the path of the ray, by accumulating colour and opacity modified by an illumination effect value at each sample point and thereby determining a colour and brightness of the pixel associated with the ray 30. In an example, the described sampling and accumulation is performed for only sample points which are within the volumetric dataset 200. That is, in this example, points along the ray 30 which do not lie within the volumetric dataset 200 do not contribute to the accumulated values determined for the ray 30.
[0042] Example methods of determining an illumination effect value for use in a method of rendering such as described with reference to Figure 3 will now be described with reference to Figure 4a and Figure 4b. In Figure 4a and Figure 4b a point
p in the volumetric dataset 200 and vectors associated with the point
p are shown for the purposes of illustrating an example local illumination model for determining an illumination effect value. The point
p is on a surface 401 which is the isosurface formed by the set of points having the same scalar value as that of the point
p. With reference to Figure 4a and Figure 4b:
N is a surface normal perpendicular to a tangent plane 402 at point p, where the tangent plane 402 is represented by a dotted line;
T is a tangent vector orthogonal to the normal N and lying within the tangent plane 402;
B is a bitangent vector orthogonal to the normal N and tangent T vectors and also lying in the tangent plane - the three vectors N, T and B form an orthonormal basis with their origin at point p;
L is a light direction vector, pointing from the point p towards an incident light source, such as light source 50 of Figure 2;
V is a viewing direction vector, pointing to a viewing point, e.g. in Figure 2 pointing to the viewpoint 10 along the direction of a ray 30;
H is a halfway vector between L and V defined as
as used in the Blinn-Phong model for computational efficiency in local illumination calculations;
φ is an azimuthal angle, describing the rotation of a vector projected onto the tangent plane 402 around the normal vector N, wherein φ_{h} is an example azimuthal angle shown in Figure 4b and is an azimuthal angle between the halfway vector H and tangent vector T ;
θ is an elevation angle, describing the angle between a vector and the surface normal N; wherein θ_{i} is an angle of elevation of the light vector L (i.e. an angle of incidence), θ_{o} is an angle of elevation of the viewing vector V (i.e. an angle of outgoing light), and θ_{h} is an angle of elevation of the halfway vector H;
wherein all vectors described above are normalised to unit length.
[0043] Referring to the scalar field represented by the volumetric dataset 200 as
s, we can define a gradient vector
g at the point
p as
and the normal vector
N is then
[0044] In some examples, an illumination effect value determined by examples methods described herein may be a value relating to a bi-directional reflection distribution function (BRDF), which is a function which models the ratio of reflected radiance leaving a surface point in a particular direction to irradiance incident on the surface point from another defined direction. The BRDF therefore may provide a ratio of outgoing light along a given viewing direction
V given a light direction
L, and the BRDF is termed herein:
f_{r} = (
L, V)
. It will be appreciated that, in examples, multiple incident light sources may be modelled, or for example, a light field may be used to model light incident on the volume 200. Therefore, in order to calculate an illumination effect value at the point
p, a sum of the contributions of the light sources present (where each may have at least one light source vector) may be determined to arrive at a final illumination effect value for the point
p.
[0045] In examples, the BRDF may be based on a local illumination model, wherein the BRDF is represented with a diffuse term and a specular term, as follows:
wherein f_{d}(L,V) is the diffuse term and k_{d} a constant weighting the diffuse term; and
wherein f_{s}(L,V) is the specular term and k_{s} a constant weighting the specular term.
[0046] Some example methods herein are methods of determining the specular term of a local illumination BRDF as formulated above. In some known examples, the specular term of the BRDF is an isotropic term. That is, it is dependent on the angles of elevation of the light vector
L and viewing vector
V to the surface normal
N, θ_{i} and
θ_{o} respectively. An isotropic specular term may also be dependent on the difference in azimuthal angle
φ between the light vector
L and the viewing vector
V, but is not dependent on the actual values of the azimuthal angles of the light vector
L and the viewing vector
V respectively.
[0047] In modelling illumination of certain types of surface, for example, brushed metal, satin, fur or vinyl or the surface of a compact disc, instead of an isotropic specular term, use of an anisotropic specular term may be appropriate, and may provide additional visual information to be imparted to a user. Examples of the present disclosure involve the use of an anisotropic illumination model, for example in a method of direct volume rendering as described with reference to Figure 2 and Figure 3. Examples of suitable anisotropic illumination models will now be described.
[0048] In some examples, a suitable anisotropic illumination model is an ellipse-based model. Here, an ellipse-based anisotropic illumination model is an illumination model which comprises one or more parameters which define a mathematical model of an ellipse. An ellipse-based anisotropic illumination model may be used to determine a specular component of a BRDF and may result in elliptically shaped specular reflection highlights when applied as an illumination effect in rendering a volumetric dataset. An ellipse-based anisotropic illumination model may, for example, comprise scalar parameters
e_{T} and
e_{B} which define lengths of major axes of an ellipse and thus the elongation of the ellipse. The orientation of the ellipse in such a model may be determined by the parameters cos
φ_{h} and sin
φ_{h}; and it should be noted that, as follows from the geometry represented by Figures 4a and 4b, cos
φ_{h} = (
H · T) and sin
φ_{h} = (
H · B). As such, the tangent vector
T, and the bitangent vector
B, in examples may effectively control the orientation of an elliptical specular highlight determined by application of an ellipse-based anisotropic illumination model.
[0049] Examples of anisotropic illumination models which comprise parameters defining an ellipse and may be used to determine an illumination effect value in methods herein are shown below.
[0050] In a first example an ellipse-based anisotropic illumination model determines a specular term of the BRDF and is defined by the following expression (1):
wherein, in expression (1),
i.e.
e_{T,B} may take the value of any of the set of real numbers not including 0. In expression (1)
F_{r} represents a Fresnel term.
[0051] In another example an ellipse-based anisotropic illumination model determines a specular term of the BRDF and is defined by the following expression (2):
wherein, in expression (2)
e_{T,B} ∈ (0,1]. That is,
e_{T,B} may be any real number greater than 0 and less than or equal to 1. In expression (2), as described above with reference to Figures 4a and 4b,
θ_{h} is the angle between the normal
N and the halfway vector
H;
θ_{i} is the angle between the normal
N and the light vector
L; and
θ_{o} is the angle between the normal
N and the viewing vector
V.
[0052] In another example, an ellipse-based anisotropic illumination model determines the specular term of the BRDF and is defined by the following expression (3):
where
a is a constant defining a shininess of a material, as may be used in a standard Phong-based reflection model. Again, in expression (3),
e_{T,B} ∈ (0,1].
[0053] Applying illumination to a sample point, for example at block 305 of the method shown in Figure 3, comprises performing a method as described according to Figure 1, further details of examples of which will be described here.
[0054] Returning to Figure 1, as mentioned above, at block 101, the example method 100 comprises applying an anisotropic illumination model, such as one of the models described above, at a given sample point by determining, based on the volumetric dataset 200, one or more parameter values relating to one or more properties of the given sample point. In examples, example anisotropic models described above comprise parameters
e_{T} and
e_{B}, and
φ_{h} (or equivalently
T and
B) for which parameter values relating to a given sample point may be determined at block 101.
[0055] Considering the first sample point 211, in some examples determining a parameter value relating to a property of the dataset 200 at the first sample point 211 comprises determining a parameter value of
e_{T} and a parameter value of
e_{B} associated with the first sample point 211. Parameter values for
e_{T} and
e_{B} may be determined for the first sample point 211 based on, for example, the volumetric dataset 200 scalar value associated with the sample point 211. For example, parameter values for
e_{T} and
e_{B} may be determined for the sample point 211 during classification of the sample point 211 by use of a transfer function. In one example, parameter values for
e_{T} and
e_{B} may be assigned for a sample point based on the opacity at the sample point, as given by the transfer function. In one example,
e_{T} and
e_{B}, may be assigned values based on an indication of the type of material the sample point 211 represents. For instance, a magnitude of the scalar value associated with the sample point 211 may be used to determine that the sample point 211 represents a bone material. A transfer function may then assign values for bone material to
e_{T} and
e_{B}, where these values may be indicative of the characteristics of anisotropic specular highlights which are representative of what may be observed on bone material. Such values may be empirically determined in some examples, for example by performing measurements of the dimensions of specular highlights observed on bone material for a particular type of lighting, or, in other examples, such values may be defined by a user of the system. In some examples, parameter values at a given sample point for input as values of ellipse parameters in an anisotropic illumination model may be determined based on other properties at the sample point. For example, a magnitude and/or orientation of the gradient of the volume at the sample point may be used to determine a parameter value for input into an anisotropic illumination model.
[0056] The values of the tangent vector
T and bitangent vector
B at a given sample point are further examples of parameter values related to a property of the volumetric dataset 200; as described above,
T and
B determine the orientation of specular highlights produced by example ellipse-based illumination models. In order to provide particular parameter values for
T and
B for input into an anisotropic illumination model, a pair of perpendicular unit vectors lying in the tangent plane at the given sample point is defined. In examples, any one of the available pairs of perpendicular unit vectors lying in the tangent plane may be defined as the parameter values for
T and
B. In some examples, the pair of unit vectors for inputting as parameter values of
T and
B may be determined by defining a fixed vector, which may for example be labelled
A, having a given orientation to the volumetric dataset 100. At a given sample point, the parameter value for
T may then in one example be defined as the cross product of
A with
N normalised to unit length; i.e.
A ×
N normalised to unit length. In this case,
B at the given sample point is assigned the parameter value
N ×
T normalised to unit length. Defining parameter values for
T and
B in this way, for input into an anisotropic illumination model, may be advantageous since it may produce illumination effects having a constant direction of anisotropy across the volume 100.
[0057] In examples of the present disclosure, one or more values relating to a curvature at a given sample point are determined as one or more parameter values relating to properties of the dataset 200 at the sample point. That is, block 101 may comprise determining one or more values relating to a curvature at the given sample point, as follows.
[0058] Returning to Figure 4a and Figure 4b, the surface 401 upon which point p lies, can be considered to have a curvature at the point p. On the surface 401, the curvature defines the change in a normal vector when applying infinitesimal positional changes on the surface 401 from the point
p. The curvature at point
p is thus based on second-order derivatives of the scalar field
s with respect to position
x,y,z at the point
p on the surface 401.
[0059] In one example, a normal plane may be defined, where the normal plane is a plane containing the normal vector
N at the point
p. It should be appreciated that a set of normal planes may be defined, each defined by an azimuthal angle of rotation
φ. The intersection of each normal plane with the surface 401, which is a two-dimensional surface, forms an intersection curve on each normal plane. From each intersection curve between the surface 401 and a normal plane, a curvature value
κ at the point
p may be determined, for example by determining a second-order derivative of the intersection curve.
[0060] In an example, by determining a curvature value for each of the normal planes, minimum and maximum curvatures for the point
p can be determined. A minimum curvature
κ_{1} and a maximum curvature
κ_{z} may collectively be referred to as principal curvatures of the surface 401 at the point
p. The principal curvatures
κ_{1} and
κ_{2} have associated with them tangential directions. The unit vectors denoting the tangential directions for the principal curvatures
κ_{1} and
κ_{2} are referred to as principal curvature directions, respectively
C_{1} and
C_{2}. It is known from differential geometry that the principal curvature directions
C_{1} and
C_{2} are always orthogonal to one another. The product of the principal curvatures
may be referred to as the Gaussian curvature, and the mean of the principal curvatures
may be referred to as the mean curvature.
[0061] It will be understood that, for volumetric data and the scalar field
s represented by that data, various methods may be used to determine principal curvatures and principal curvature directions. For example, methods described in the work of
Kindlemann et al. (Curvature-based transfer functions for direct volume rendering: Methods and applications, In Visualization, 2003. VIS 2003. IEEE (pp. 513-520). IEEE) may be used to reconstruct principal curvatures and principal curvature directions from second-order partial derivatives of the scalar field. In some examples curvature values may be determined based on a smoothed version of the volumetric dataset 200. For example, smoothing of the volume may be performed prior to calculation of curvature values via use of a spatial smoothing filter such as a box filter, a tent filter or a Gaussian filter. Furthermore, example methods of determining curvature values for use in methods described herein may define a limit to the range of permitted curvature values. For example, in order to reduce variance in curvature measurements, principal curvature values may be limited to a range within a number k of standard deviations around the mean of the curvature values determined for the dataset. Here, k may, for example, be 1, 2 or 3.
[0062] At block 102 parameter values related to the given sample point are provided as inputs to the anisotropic illumination model being used. In some examples, parameter values determined at block 101 and relating to the given sample point are input into the anisotropic illumination model being used as values of parameters
e_{T} and
e_{B}. In some examples, as described above, the parameter values determined for
e_{T} and
e_{B} for the given sample point are not related to curvature at the given sample point, instead being determined by a transfer function or the like. In some examples, however, parameter values determined at block 101 and relating to curvature at the sample point are provided as values of each of
e_{T} and
e_{B}.
[0063] When determining an example parameter value for input into one of the above anisotropic illumination models, to account for the different ranges of
e_{T} and
e_{B} used in said models, a normalisation factor
n may be defined. The normalisation factor
n acts to map the range of determined parameter values to a range suitable for use as values of the parameters
e_{T} and
e_{B} in the given anisotropic illumination model being used.
[0064] Examples of parameter values which may be input as values of
e_{T} and
e_{B} will now be described. The below examples may be considered to define different mappings of curvature values to the ellipse-based values present in example anisotropic illumination models.
[0065] In one example, parameter values defining a locally normalised curvature at the given sample point are input as values of parameters
e_{T} and
e_{B}. That is, an anisotropic illumination model may have parameters assigned parameter values as follows:
and
wherein
is the unit vector formed from the absolute values |
κ_{1}| and |
κ_{2}| of the two principal curvatures
κ_{1} and
κ_{2} at the sample point.
[0066] That is, in this example, the parameters defining the elongation of the ellipse in the particular anisotropic illumination model being applied in this example are assigned values dependent on the product of the normalisation value n and the components,
and
of the unit vector formed from the absolute values |
κ_{1}|, |
κ_{2}| of the principal curvatures at the given sample point.
[0067] As mentioned above,
n is the above-described normalisation factor for defining the range of values which may be input into a given anisotropic illumination model as values of the parameters
e_{T} and
e_{B}. For example, when applying the model defined by expression (1) to determine a specular term at a sample point, i.e. determining the value of
f_{s,1} at the sample point,
n has a value which may be greater than 1, because, in the expression for
f_{s,1},
However, when applying either the model defined by expression (2) or the model defined by expression (3) to determine the specular term at the sample point i.e. determining a value for
f_{s,2} or for
f_{s,3},
n = 1 because in these models
e_{T,B} ∈ (0,1].
[0068] In another example, parameter values for input as values of the parameters
e_{T} and
e_{B} may be an inverse of locally normalised principal curvatures at the sample point. That is, parameter values may be assigned as follows:
where
and
are the locally normalised principal curvatures defined as above.
[0069] In another example, a parameter value for input as one of the values of
e_{T} and
e_{B} may be a ratio of the locally normalised principal curvatures at the sample point. In examples, the other parameter value for input to the other of
e_{T} and
e_{B} may be normalisation value
n. That is, parameter values may be assigned as follows:
[0070] In another example, a parameter value for input as
e_{T} and
e_{B} may be Gaussian curvature calculated from the locally normalised principal curvatures at the given sample point, while the other parameter for input may be the normalisation value
n. That is:
where
and
is the Gaussian curvature; and (
κ̂_{1},
κ̂_{2}) is the normalised unit vector formed from signed values of principal curvatures, i.e.
[0071] In another example a parameter value for input as values of parameters
e_{T} and
e_{B} may be a mean curvature, i.e. the mean of the components of the unit vector of locally normalised signed curvature, while the other parameter value for input may again be the normalisation value n. That is:
where
[0072] In examples, parameter values determined at block 101 for vectors
T and
B may also be input at block 102. It should be noted that in practice this may mean selecting the orientation of the orthonormal basis used at the sample point, i.e. selecting a tangent vector
T and a bitangent vector
B. In practice, a parameter value input into a model may be a parameter value for
φ_{h}, which is dependent on
T and
B and appears in the above example anisotropic illumination models.
[0073] In some examples, a parameter value for the tangent vector
T is selected, for example arbitrarily, from the set of possible tangent vectors lying in the tangent plane at the sample point. Conversely, a parameter value for the bitangent vector
B, orthogonal to the tangent vector
T may be selected and the tangent vector
T uniquely defined by this selection.
[0074] In examples, parameter values for input into an anisotropic illumination model as values of parameters
T and
B relate to the curvature of the volumetric dataset 200 at the sample point. For example, the principal curvature directions
C_{1} and
C_{2} (which are unit vectors) at the sample point may be determined as described above and input into a model as values of the parameters
T and
B.
[0075] In one example, principal curvature direction
C_{1} may be input as a parameter value for the tangent vector
T, i.e.
T =
C_{1} and
C_{2} may be input as a parameter value for the bitangent vector
B, i.e.
B =
C_{1}. As such, the orientation of the specular highlight produced by an example anisotropic illumination model may be configured to follow the principal curvature directions at the given sample point.
[0076] As can be seen from the above, in examples according to the present disclosure, parameter values relating to the curvature at a sample point at which an illumination effect is being applied may be input into an anisotropic illumination model being used, to thereby determine an anisotropic specular illumination effect value at the point. The illumination effect value, as described above is used to modulate the optical properties of the given point for accumulating into the sum of the values of the optical property along the ray being simulated. Therefore, example methods may achieve an anisotropic illumination effect which is dependent on the curvature of a surface at the given sample point. This may provide additional visual information to a viewer and enable a viewer to identify curved portions of an object, such as the object 210, within the volumetric dataset 200. In turn this may provide for increased realism in a rendered image of the volumetric dataset 200 provided to the viewer or improved perception by the viewer of, for example, boundaries, shape and depth within the volumetric dataset 200 from the resulting image.
[0077] In example methods of determining an illumination effect value during direct volume rendering, an anisotropic illumination model is selected, and one or more parameter values for inputting as values of parameters of the selected anisotropic illumination model are also selected. A selected anisotropic illumination model may be, for example, one of the above described ellipse-based models. Similarly, the one or more parameter values may be, for example, values as described above to be input as the values of parameters defining an ellipse and forming part of the selected illumination model. For example, the method 100 may, in some examples, comprise selecting a particular anisotropic illumination model for determining an illumination effect value to be applied at a particular sample point in the volumetric dataset 200. It will be appreciated that in some examples, the same anisotropic illumination model could be used at each sample point, while in other examples a different anisotropic illumination model could be selected to be applied at different sample points within the volumetric dataset 200. In yet other examples, an anisotropic illumination model may only be applied at selected sample points within the dataset 200. For example, an anisotropic illumination model may be applied at points in the dataset 200 where it is determined that an anisotropic illumination effect may be beneficial to a viewer's perception of that point. In such an example, a different illumination model, for example an isotropic local illumination model may be applied to determine illumination effect values at other points within the dataset 200, for example where anisotropic illumination effects are considered less likely to be of importance to a viewer's perception.
[0078] Similarly, selected parameter values for inputting into a selected anisotropic illumination model may differ for different sample points within the dataset 200. As such, the method 100 may comprise selecting a particular mapping of parameter values for inputting as values of parameters in the illumination model being used. Notably, any of the above options for controlling ellipse elongation, e.g. for inputting as parameter values of parameters
e_{T} and
e_{B}, may be combined with any of the above options for controlling ellipse orientation, e.g. for providing input values for
T and
B. In some examples, the assignment of parameter values to parameters of a given anisotropic illumination model may be swapped. That is, for example, the above expressions for
e_{T} and
e_{B} may be reversed such that, for example
e_{B} and
e_{T} =
Similarly, the parameter values input as values of
T and
B may be swapped, such that, in one example
T = C_{2} and
B = C_{1}.
[0079] Selection of an anisotropic illumination model and/or a selection of parameter values and how they are input into the selected anisotropic illumination model at a particular sample point. i.e. how they are mapped to parameters of the selected illumination model, may be dependent on, for example, local criteria defining the sample point or its surrounding in the dataset 200. Examples of such criteria are surface properties, occlusion, depth, material, opacity, segmentation etc.
[0080] Referring now to Figure 5, there is illustrated schematically an example system 501 in which an example rendering apparatus 504 may be used. The system 501 comprises a scanner 502, the rendering apparatus 504, and a visualisation unit 514. In some examples, the system may comprise fewer components than or additional components to those illustrated in Figure 5. For example, the system 501 may comprise a computer network such as the internet.
[0081] The scanner 502 may be any scanner for generating a dataset comprising the volumetric dataset 200, which, as described may, for example, be a medical volumetric dataset representing a portion of a patient. For example, the scanner 502 may be a computed tomography (CT) scanner, a magnetic resonance imaging (MRI) scanner, a positron emission tomography (PET) scanner, an ultrasound scanner or the like. In another example the scanner 502 may, for example, be for producing a volumetric dataset representing geological data. The scanner 502 is connected to the rendering apparatus 504, for example via wired or wireless connection. The scanner 502 may be arranged to provide the volumetric dataset to the rendering apparatus 504.
[0082] The rendering apparatus 504 comprises a processor 506 and a memory, in the form of a storage 508. In this example, the rendering apparatus 504 is arranged to perform the above described method of determining an illumination effect and/or of rendering the volumetric dataset. For example, the storage 508 may comprise a machine-readable medium comprising a set of machine readable instructions which when executed by the processor 506 cause the rendering apparatus 504 to perform an above-described example method. The program may be stored on a computer readable medium which may be read by the rendering apparatus 504 to thereby execute the program. The rendering apparatus 504 may be arranged to receive directly or indirectly or otherwise acquire from the scanner 502 the volumetric dataset 200.
[0083] The volumetric renderer 504 may comprise a processor for operating any rendering algorithm capable of simulating light transport within the volumetric dataset 200. An example suitable rendering algorithm is a ray casting method comprising a method of determining an illumination effect value as described above.
[0084] The rendering apparatus 504 may be arranged to transmit information, for example, a colour value for each pixel in a 2D image plane, to a visualisation unit 514. The transmission may be direct or indirect, for example via a wired connection, a wireless connection, or via the internet.
[0085] The visualisation unit 514 may comprise visualisation software for displaying a two-dimensional projection of a three-dimensional object 510. The visualisation unit 514 may comprise a display screen, and one or more graphics hardware or software components. In some examples, the visualisation unit 514 may be or comprise a mobile device. In some examples the visualisation unit 514 may comprise a virtual-reality device.
[0086] Although the invention has been described in the context of a direct volume rendering algorithm employing a ray casting approach, it should be appreciated that the invention may be applied in other example methods of visualising a volume. For example, any volume rendering method which comprises determining an illumination effect value may use a method of determining an illumination effect value by applying an anisotropic illumination model as described herein. For example, methods of determining an illumination effect described herein may be employed in volume rendering techniques such as path tracing, splatting, or shear warp.
[0087] The above embodiments are to be understood as illustrative examples of the invention. Other embodiments are envisaged. It is to be understood that any feature described in relation to any one embodiment may be used alone, or in combination with other features described, and may also be used in combination with one or more features of any other of the embodiments, or any combination of any other of the embodiments. Furthermore, equivalents and modifications not described above may also be employed without departing from the scope of the invention, which is defined in the accompanying claims.