Image capture and processing

A method comprising:

• receiving an albedo image showing human skin;
• receiving or determining spectral data corresponding to an illumination condition used to obtain the albedo image;
• processing the albedo image to fit parameter maps for a spectral bidirectional scattering surface reflectance distribution function skin model, the parameter maps comprising a modelled melanin concentration, a modelled haemoglobin concentration, a modelled melanin blend-type fraction and a modelled epidermal haemoglobin fraction;
• wherein at least three of the parameter maps are independent;
• wherein processing the albedo image to fit parameter maps comprises applying a neural network model to inputs comprising:
• the albedo image;
• and
• the spectral data;
• wherein a final stage of the neural network outputs an output RGB albedo image, and wherein the neural network is configured to determine the parameter maps which minimise differences between the output RGB albedo image and the albedo image.

A method of image processing, comprising:

• receiving a first image of human skin, the first image corresponding to a first, uniform broadband illumination condition;
• receiving a second image which has the same field of view and contents as the first image, the second image corresponding to a second illumination condition which comprises a uniform narrowband illumination condition;
• receiving or determining first spectral data corresponding to the uniform broadband illumination condition and second spectral data corresponding to the second illumination condition;
• processing the first and second images to fit parameter maps for a spectral bidirectional scattering surface reflectance distribution function skin model, the parameter maps comprising a modelled melanin concentration, a modelled haemoglobin concentration, a modelled melanin blend-type fraction and a modelled epidermal haemoglobin fraction;
• wherein at least three of the parameter maps are independent;
• wherein processing first and second images to fit parameter maps comprises using a three-dimensional spectral look-up table or using a four-dimensional spectral look-up table;
• wherein the spectral look-up table is constructed based on spectral information comprising the first spectral data and the second spectral data, and further based on the spectral skin reflectance model.

A method comprising:

• receiving an albedo image showing human skin;
• receiving or determining spectral data corresponding to an illumination condition used to obtain the albedo image;
• processing the albedo image to fit parameter maps for a spectral bidirectional scattering surface reflectance distribution function skin model, the parameter maps comprising a modelled melanin concentration, a modelled haemoglobin concentration, a modelled melanin blend-type fraction and a modelled epidermal haemoglobin fraction;
• wherein at least three of the parameter maps are independent;
• wherein processing the albedo image to fit parameter maps comprises using a three-dimensional spectral look-up table or using a four-dimensional spectral look-up table;
• wherein the spectral look-up table is constructed based on the spectral data, and further based on the spectral skin reflectance model.

A method comprising:

• receiving an albedo image showing human skin;
• receiving or determining spectral data corresponding to an illumination condition used to obtain the albedo image;
• receiving a first input parameter map related to melanin concentration and corresponding to the albedo image;
• receiving a second input parameter map related to haemoglobin concentration and corresponding to the albedo image;
• processing the albedo image to fit parameter maps for a spectral bidirectional scattering surface reflectance distribution function skin model, the parameter maps comprising a modelled melanin concentration, a modelled haemoglobin concentration, a modelled melanin blend-type fraction and a modelled epidermal haemoglobin fraction;
• wherein at least three of the parameter maps are independent;
• wherein processing the albedo image to fit parameter maps comprises:
• processing the albedo image to fit first and second intermediate parameter maps for a two parameter spectral bidirectional scattering surface reflectance distribution function skin model, wherein the first intermediate parameter map comprises an intermediate melanin concentration and the second intermediate parameter map comprises an intermediate haemoglobin concentration, wherein the intermediate parameter maps are determined using a first two-dimensional spectral look up table constructed based on the spectral data;
• re-scaling the first and second input parameter maps based on the intermediate parameter maps;
• setting the parameter map corresponding to the modelled melanin concentration equal to the re-scaled first input parameter map;
• setting the parameter map corresponding to the modelled haemoglobin concentration equal to the re-scaled second input parameter map;
• processing the albedo image to fit the parameter maps corresponding to the modelled melanin blend-type fraction and the modelled epidermal haemoglobin fraction, wherein the parameter maps are determined using a second two-dimensional spectral look up table constructed based on the spectral data.