Stereoscopic display system

Abstract

A method of rendering a stereoscopic image 340 is described where in the stereoscopic image comprises a left image and a right image. The method comprises initialising a computer model of the object 2

Description

TECHNICAL FIELD The invention relates to a method of and a system for rendering a stereoscopic image (or frame) of an object for displaying on a stereoscopic display device (sometimes referred to as a 3D display) to an observer viewing the stereoscopic display device. BACKGROUND Stereoscopic microscopes are optical devices which generally employ two separate optical paths with two objectives (e.g., a left and a right objective) and eyepieces to provide two different viewing angles to the left and right eyes of an observer viewing the stereoscopic microscope. This produces a three-dimensional visualization of the object being examined. Stereoscopic microscopes are often used for microsurgery or robotic assisted surgery. Example stereoscopic microscopes also include slit lamp microscopes used during an eye examination. The slit lamp microscope may be part of a stereoscopic microscope system, incorporating a processor to provide an electronic display of the object and to record images for future reference. For example, it may be beneficial to be able to also provide the three-dimensional visualisation of the object on a stereoscopic display device to observers, such as to students as a teaching aid and other surgical staff or, in the case of robotic assisted surgery, to the surgeon themselves. A stereoscopic display device projects different images to the left and right eyes to its observer, such that the observer can perceive the depth of the three-dimensional visualisation through binocular disparity. It has also been realised that it is beneficial to provide a three-dimensional visualisation of the object viewable from a range of positions, such that the observer is not spatially restricted in order to view the three-dimensional visualisation but rather the observer can move relative to the three-dimensional visualisation. For instance, with stereoscopic microscopes the observer has to fixedly position themselves to look through the eyepieces. The ability to move relative to the visualisation can improve their ease of viewing the object as well as potentially allows the observer to inspect the object from different perspectives. Furthermore, being able to move relative to the three-dimensional visualisation can improve the ability of the observer to perceive depth, and hence detail, from it. It may allow an observer to notice a feature which can then be inspected more closely using the microscope optics. This is particularly beneficial in stereoscopic microscope systems, where the optics are typically fixed such that the only way to achieve this effect is by manipulating the object being viewed. However, this is often impractical to do so due to the small scale of the object being viewed and is further complicated where the object being viewed is that of a patient. For instance, where the object being viewed is an eye of a patient during an eye examination, the patient’s head is supported relative to the slit lamp microscope by a head support to keep it as still as possible, with only the illumination being movable horizontally across the eye of the patient to highlight features. To provide a visualisation of the object on a display, a portion of the light from each of the two objectives is typically diverted to image sensors (e.g., CCD or CMOS sensors), e.g., using a pair of beam splitters, one for each optical path. The signal from the image sensors can then be directed to a display for a user to observe the visualisation of the object. Conventionally this would be a two-dimensional screen display, wherein the signals from the image sensors would be displayed side by side. Some improvements have been made with three-dimensional projector systems though these are bulky, relatively expensive and require a reasonably precise positioning of the observer’s head to ensure that different images are projected into each of the observer’s eyes. SUMMARY According to a first aspect, there is provided a method of rendering a stereoscopic image of an object for displaying on a stereoscopic display device to an observer viewing the stereoscopic display device. The stereoscopic image comprises a left image and a right image. The method comprises receiving a first image of the object from a first image sensor, receiving a second image of the object from a second image sensor, receiving head tracking data of the observer, determining one or more render camera viewpoints relative to the computer model based on the head tracking data, and rendering the stereoscopic image based on a computer model and a texture mapping using the one or more render camera viewpoints. The left image of the stereoscopic image is rendered using the first image as the texture and the right image of the stereoscopic image is rendered using the second image as the texture. The method may be executed on a data processing device, e.g., a processor, of a computer. The method may comprise initialising (on the data processing device of a computer) a computer model of the object and initialising a texture mapping for rendering the computer model with a texture. Following the initialising, the method may comprise receiving a first image of the object from a first image sensor, receiving a second image of the object from a second image sensor, receiving head tracking data of the observer, determining one or more render camera viewpoints relative to the computer model based on the head tracking data, and rendering the stereoscopic image based on the computer model and the texture mapping using the one or more render camera viewpoints. Again, the left image of the stereoscopic image is rendered using the first image as the texture and the right image of the stereoscopic image is rendered using the second image as the texture. The receiving of the first and second images may comprise capturing, by the data processing device (e.g., a processor), the first and second images from a signal output of the first and second image sensors. The receiving of the first and second images may comprise loading, by the data processing device, the first and second images from a local storage device and / or a remote database. These loaded first and second images may have been originally captured by the first and second image sensors and saved to the local storage device and / or remote database for later user in the method. As such, the first and image sensors may be located remotely to where the method of rendering the stereoscopic image is executed. When repeating the method to render a new stereoscopic image, the initialisation of the computer model may not be repeated but instead the previously initialised computer model may be used. The method of rendering a stereoscopic image produces a stereoscopic image which, when displayed on a stereoscopic display device, provides a three-dimensional visualisation of the object. The stereoscopic image is rendered based on one or more render camera viewpoints that is determined based on head tracking data. The head tracking data may comprise data indicative of eye positions. The stereoscopic image may be viewable from different head positions in space through head tracking functionality available from the stereoscopic display device. This can be beneficial as it does not require the observer to maintain a still central position. An observer wishing to inspect the three-dimensional visualisation from a different perspective, may be able to move their head and the step of rendering the stereoscopic image in the above methods can be repeated with different head tracking data representing the new (different) desired perspective. The left image may be for displaying to a left eye of an observer, and the right image may be for displaying to a right eye of the observer. The method of rendering an image can offer advantageous functionality to optical microscopes, like a slit lamp microscope, which is being used to observe a subject’s eye, where the stereoscopic display device is being used in parallel to observe (simultaneously) the normal optical output which would be seen viewing the subject’s eye through the binocular viewer of the slit lamp microscope. The method is not limited to slit lamp microscopes and could be used in conjunction with other types of microscope or other visual inspection instruments having a stereoscopic optical output. However, the method is particularly well suited to a slit lamp microscope application and especially where the stereoscopic display device comprises a screen with a micro optical lens array (e.g., a lenticular lens comprising an array of lenses) that are configured to direct a left image to a left eye of an observer and a right image to a right eye of the observer. In such an arrangement, the stereoscopic image can be seen as a three-dimensional visualisation of the subject’s eye, in real-time, from a wide range of viewing angles (e.g., more than 60°, i.e., more than 30° on each side of an optical axis) without a need for special glasses or headgear, allowing the stereoscopic display device to be viewed easily for quickly inspecting an area of the object or for assisting with training a student. The method of rendering an image may be used in generating a stereoscopic video for display on the stereoscopic display device to allow the observer (which might be a student) to watch an examination of the subject’s eye, in real-time, while a practitioner (e.g., a qualified ophthalmologist) controls the examination with the slit lamp microscope (e.g., controlling the direction of the illumination, the aperture and the magnification settings, etc.). Rendering the stereoscopic image may require an image plane, which may depend upon the method of rendering, onto which the stereoscopic image is rendered. The image plane may be determined based on a position of the stereoscopic display device. The first and second images of the object are different. The first and second images of the object may be at different viewing angles. Stereopsis is a component of depth perception retrieved by means of binocular disparity through binocular vision, i.e. through the differences in the left and right images of the stereoscopic image which may be displayed to the left and right eyes of the observer respectively. This method utilizes any existing binocular disparity between the first and the second images of the object by rendering the left and right images based on the first and second images respectively. This is in comparison to other methods which may try to combine first and second images to use as a single texture for rendering both the left and right images of the stereoscopic image, which can be computationally more complex and lose detail from the original images. The texture map may be any suitable mapping such as a UV map, wherein each vertex in a polygon of the computer model is assigned a texture coordinate in a texture. A texture is an image, typically a 2D image, containing a colour descriptor for each coordinate / pixel. The colour descriptor can include one or more of RGB data, opacity, etc. Optionally in the first aspect: the method is used in conjunction with a slit lamp microscope which comprises the first and second image sensors, the object is an eye, and the first image sensor is spaced apart from the second image sensor in a stereo camera arrangement. The stereo camera arrangement may also be known as a stereoscopic camera, and these terms are used interchangeably throughout this application. The stereo camera arrangement, for example, may comprise a pair of beam splitters, e.g., mounted securely within a housing, each beam splitter deflecting a portion of light in a left or right side of the microscope toward an image sensor. The stereo camera arrangement may comprise a bayonet fitting on one side for attachment to a microscope body and a bayonet fitting on an opposed side for attachment of a binocular viewer. The method may advantageously be used with any slit lamp microscope comprising first and second image sensors. The first and second image sensors may be arranged opposite each other, on opposite sides of the microscope. The first and second image sensors may be spaced apart from each other, i.e., they are different image sensors. Their viewing angles of the object may diverge by a small amount. The small amount may be less than 15 degrees, for example, less than 10 degrees. The small amount may be more than 2 degrees, for example, 3 degrees or more, may be 5 degrees or more. This viewing angle may be measured between an optical axis of the first image sensor and an optical axis of the second image sensor. The slit lamp microscope may comprise a binocular viewer for an observer, e.g., a second observer, to view the object; and the stereo camera arrangement may be mounted between a microscope body and the binocular viewer of the slit lamp microscope. The binocular viewer and the stereo camera may be arranged to capture the same field of view at the same magnification. The binocular viewer and the stereo camera may be arranged to capture the same views of the object. The head tracking data of the observer may be relative to the stereoscopic display device. The head tracking data may comprise a pair of eye positions. The head tracking data may comprise a left eye position and a right eye position. The head tracking data may be output directly from the display device, for example, as a parameter which can describe a position of a head in physical space. The display device may be interrogable with a software command to deliver the head tracking data, as and when requested. This could be on the basis of a frame interval from a video feed coming from one or more of the first and / or second image sensors, e.g., 30 times a second or faster. The computer model of the object may be a pre-determined computer model. A pre-determined computer model may be understood as an a-priori model of the object being viewed, i.e. not based on any observed information regarding the object but instead deduced based on theoretical deduction. In particular, the predetermined model may be based on characteristics common to the object being viewed. The pre-determined computer model may be a plane. That is, a flat surface. The plane may be a bounded plane. The plane may have set dimensions to bound the plane. Boundaries of the plane may correspond to an aspect ratio of the first and second images and comprise a plurality of vertices. The number of vertices may be proportional or equal to the number of pixels in each of the first and second images. The distribution of vertices within the bounded plane may be proportional or equal to the distribution of pixels in each of the first and second images. While a plane minimises computational load and provides an acceptable stereoscopic image, the pre-determined model may also be a curved surface. The curvature may be in one dimension or more than one dimension, for example, corresponding in one or more dimensions to curvature of a retina. Using a pre-determined computer model advantageously reduces the compute time to render the stereoscopic image. This also can reduce the computational power necessary, thereby reducing the equipment necessary to carry out the method at a certain speed. It has been determined that a pre-determined computer model can be a sufficient approximation of the object to achieve an inspectable three-dimensional visualisation having acceptable accuracy. This is because, as mentioned above, the three-dimensional representation takes advantage of the existing binocular disparity between the first and the second images of the eye by rendering the left and right images of the stereoscopic image based on the first and second images respectively. For example, where the object is an eye being viewed through a slit lamp microscope, it has been found that a computer model in the form of a plane offers good visualisation of the eye while minimising computational complexity. There is minimal observable lag. The computer model may be initialised with a pre-determined computer model and the method may involve refining the pre-determined computer model based on the first and / or second images, and / or other observed data regarding the object. Refining the pre-determined computer model may involve estimating a depth map based on the first and second images, and updating parameters of the computer model to fit it to the depth map. The method may comprise receiving gesture information of the observer. The gesture information may be indicative of a desired interaction with the computer model. The desired interaction may be one or more of adjusting a render camera viewpoint relative to the computer model, measuring a feature on the computer model, marking a feature on the computer model, etc. The head tracking data of the observer may comprise a pair of eye positions of the observer; and the one or more render camera viewpoints may be a pair of render camera viewpoints. The method may comprise: rendering the left image of the stereoscopic image based on a left render camera viewpoint of the pair of render camera viewpoints; and rendering the right image of the stereoscopic image based on a right render camera viewpoint of the pair of render camera viewpoints. The method may comprise: determining a separation plane based on the head tracking data of the observer, wherein the separation plane corresponds to a sagittal plane of a head of the observer; and determining which of the pair of render camera viewpoints correspond to the left render camera viewpoint and the right render camera viewpoint based on which side of the separation plane each of the pair of render camera viewpoints lies. The sagittal plane of a human is an anatomical plane that divides the body into right and left sections. In this case, the sagittal plane may be that of the head of the observer and mid-sagittal plane. This determination of a separation plane may be performed because in some instances the head tracking data may be unlabelled, such that it is not possible to directly determine which of the pair of render camera viewpoints correspond to the left eye and right eye of the observer. However, the head tracking data may enable the sagittal plane to be determined. By referencing the sagittal plane, it is possible to easily determine which of the pair of render camera viewpoints correspond to the left eye and right eye of the observer based on which side of the plane they lie. According to a second aspect, there is provided a method of rendering a stereoscopic video for displaying on a stereoscopic display device to an observer viewing the stereoscopic display device, wherein each stereoscopic image of the stereoscopic video is rendered according to the method of the first aspect. When rendering the stereoscopic video, the step of initialising the computer model may only be performed once for a first stereoscopic image of the stereoscopic video. The first and second images may be images of a first and a second video, wherein the first and second images may be time synced. The first and the second videos may be livestreams. The head tracking data may be in real-time and the stereoscopic video may be a livestream for displaying on the stereoscopic display device to the observer in real-time. Where the computer model is initialised with a pre-determined computer model and refined based on the first and / or second images of the first and / or second video, the refining may be performed each time the stereoscopic image of the stereoscopic video is rendered or at intervals. Where an observer can view the livestream in real-time, this enables them to perform actions away from the slit lamp microscope (or other microscope) as they no longer need to look through a binocular viewer of the slit lamp microscope in order to view the object. For instance, this may aid in studying the patient (i.e., the subject), explaining an observation to a student, or even enable the observer to perform a medical intervention without being unduly constrained. The method according to the first or second aspect may comprise displaying the stereoscopic image or stereoscopic video on the stereoscopic display device to the observer. The method may comprise saving the stereoscopic image or stereoscopic video electronically. The stereoscopic image or stereoscopic video may be saved on a local storage device and / or a remote database, optionally with clinical features and / or a diagnosis of the object. This can be retrieved later for subsequent viewing on the stereoscopic display device. The saved stereoscopic images / videos may provide a dataset that can be searched. The library may be searched automatically by background software running locally on the data processing device or on a cloudbased system using feature-based recognition software (e.g., which may rely on traditional computer vision techniques or on machine learning based solutions such an involving convolutional neural networks (CNN) which have been trained for image detection relating to the object being observed, for example, defects in eyes). Where a defect is identified or a diagnosis is made automatically by the software, then a notification may be displayed or otherwise conveyed to the observer, e.g., on the display device, to bring the defect or diagnosis to the attention of the observer for the observer to investigate further. Such identification and diagnosis processes may be run automatically in parallel in the background during normal operations being performed by the observer, to assist the observer in identifying and diagnosing defects during an examination. The stereoscopic image or stereoscopic video may be saved, optionally with clinical features and / or a diagnosis of the object, in a dataset. The dataset may be used with feature detection software to assist with identification and diagnosis processes. The dataset may be suitable for machine learning. For example, the dataset may be used to generate a deep neural network that, based on an inputted stereoscopic image or video, predicts the diagnosis via classification and highlights clinical features via semantic segmentation of the inputted stereoscopic image or video. However, such machine learning requires labelled training data (e.g., a diagnosis and / or semantic segmentation labelling of the image by a qualified ophthalmologist). Therefore, unsupervised, semi-supervised learning techniques, and / or a combination of these are equally contemplated. The stereoscopic image or video may also be shared with other users with access to such a stereoscopic display device. For example, a stereoscopic display device may be located remotely to the slit lamp microscope, and an observer may be a remote observer. The stereoscopic display device may communicate over a network to the slit lamp microscope. This can enable a remote observer, such as a more-experienced specialist or remote student, to view a three-dimension visualisation of the object on the stereoscopic display device. There may be a plurality of stereoscopic display devices, wherein the stereoscopic image or video is displayed on each of the stereoscopic display devices. This can enable a plurality of observers to view the object on the stereoscopic display devices, for example, allowing an examination to be broadcast to a plurality of students or other observers simultaneously. While the stereoscopic display device is preferably a table or counter surface mounted screen device, or similar (e.g., a movable mounting arm or wall-mounted device), the stereoscopic display device could be a virtual reality headset. The stereoscopic display device may comprise: a pixel array comprising a plurality of pixels for displaying the stereoscopic image; a micro optical lens array arranged over the pixel array and comprising a plurality of micro optical lenses, wherein each micro optical lens is arranged over multiple pixels of the pixel array and configured to project light from the multiple pixels over different viewing angles; and the head tracking data of the observer comprises a pair of eye positions of the observer, wherein the pair of eye positions is a left eye position and a right eye position. An advantage of certain stereoscopic display devices, such as one comprising micro optical lenses, is that they permit the observer to view the stereoscopic image without the observer needing to wear any special equipment such as glasses (glasses free). However, for multiple observers to view the same stereoscopic display device, it may require a certain type of stereoscopic display device and observers to wear special equipment such as glasses. The plurality of micro optical lenses may be a lenticular lens comprising a plurality lenses. Each micro optical lens may be arranged over at least a left pixel and a right pixel of a pixel pair to direct light from a left pixel generally in a first direction and light from a right pixel generally in a second direction different to the first direction, such that the light from the left pixel and the right pixel diverge from the micro optical lens(es) by an angle where an amount of divergence generally corresponds to an average separation of left and right eyes of an observer at an expected viewing distance from the stereoscopic display device. The stereoscopic display device may comprise multiple columns of such pixels arranged vertically with respect to the stereoscopic display device. The lenses of the lenticular lens may each have a convex profile in a horizontal direction, e.g., in a direction extending across a pair of left and right pixels, and extend prismatically in a vertical direction with respect to the stereoscopic display device (albeit with the stereoscopic display device may be inclined at 45° for the observer to view the three-dimensional representation easily). The plurality of micro optical lenses may be designed so that, at different viewing angles, different pixels of the pixel array are visible through each micro optical lens. The method may comprise: interlacing the left image and the right image of the stereoscopic image; displaying the interlaced stereoscopic image on the pixel array; and projecting, with the micro optical lens array, the left image towards the left eye position, and the right image towards the right eye position. The projecting may be in the sense of directing a maximum amplitude direction of light in a given direction. Light from the pixels may be viewable to an extent from other directions but this would be much fainter and generally ignored in the mind of the observer. The projecting of the left and right images to the left and right eye positions respectively may involve determining, out of the multiple pixels arranged under each micro-optical lens, which pixels is / are visible to the left eye position only and which is / are visible to the right eye position only. The interlacing and displaying of the interlaced stereoscopic image may be performed such that the left image is displayed on pixels visible to the left eye position only and the right image is displayed on pixels visible to the right eye position only. The method of rendering the video may include outputting a video feed comprising left image and right image components of the interlaced stereoscopic image in a sequence which is readable by the stereoscopic display device, the sequence causing left pixels of pixel pairs to display left image components of the interlaced stereoscopic image intended for the observer’s left eye, and causing right pixels of the pixel pairs to display right image components of the interlaced stereoscopic image intended for the observer’s right eye. According to a third aspect, there is provided a method of rendering a stereoscopic image of an object for displaying on a stereoscopic display device to an observer viewing the stereoscopic display device, the stereoscopic image comprising a left image and a right image, the method comprising: receiving a first image of the object from a first image sensor; receiving a second image of the object from a second image sensor, wherein the first image sensor is spaced apart from the second image sensor in a stereo camera arrangement; receiving head tracking data of the observer from the stereoscopic display device, wherein the head tracking data describes a pair of eye positions of the observer; determining a pair of render camera viewpoints relative to a computer model based on the pair of eye positions; rendering each pixel of the stereoscopic image based on the computer model and a texture mapping, the rendering using the head tracking data to select a texture to apply with the texture mapping, wherein: if the pixel is determined to be of the left image of the stereoscopic image, the pixel is rendered using the first image as the texture and a left camera viewpoint; and if the pixel is determined to be of the right image of the stereoscopic image, the pixel is rendered using the second image as the texture and a right camera viewpoint. The method according to the third aspect may comprise: determining a separation plane based on the head tracking data of the observer, wherein the separation plane corresponds to a sagittal plane of a head of the observer; and determining which of the pair of render camera viewpoints correspond to the left render camera viewpoint and the right render camera viewpoint based on which side of the separation plane each of the pair of render camera viewpoints lies. The method may be a method of rendering stereoscopic images to generate a video for display on the stereoscopic display device. The stereoscopic video may comprise left and right images that are paired as frames of the video. The left and right images may be time synced. The left and right images may be configured for observing on a lenticular stereoscopic display device comprising a micro optical lens array arranged over a pixel array. The video may comprise at least 30 frames per second. The first and second image sensors may be part of a slit lamp microscope system and they may be arranged to feed (output) live video feeds to a data processing device where the head tracking data is analysed, the camera viewpoints are determined and the computer model and texture mapping are initialised and used for rendering the stereoscopic image. Thus, viewed from a fourth aspect there is provided a method of generating a video feed for a stereoscopic display device, the video feed for displaying a three-dimensional representation of an object to an observer viewing a screen of the stereoscopic display device. The method comprises: receiving a first video signal (feed) from a first image sensor positioned to observe a first optical output of an object at a first viewing angle, the first video signal comprising a sequence of first frames; receiving a second video signal (feed) from a second image sensor positioned to observe a second optical output of the object at a second viewing angle different to the first, the second video signal comprising a sequence of second frames; processing first frames and corresponding second frames as frame pairs; reading a head tracking parameter (head tracking data) from the stereoscopic display device that is indicative of left and right eye position for the observer, the parameter being output in real time from the stereoscopic display device and read for at least each frame pair; and generating a video feed of stereoscopic image frames for the screen of the stereoscopic display device, wherein the screen comprises a lenticular display of lenses arranged over an array of pixels, the lenses being configured to display, in use, a first image to a left eye of the observer and a second image to a right eye of the observer, wherein the head tracking parameter from the stereoscopic display device is used in a rendering process to automatically select either the first frame or the second frame of each frame pair as basis when rendering each pixel of the display. Each lens may be arranged over a plurality of pixels in a horizontal direction of the screen and configured to display each of the plurality of pixels over different fields of view. The head tracking parameter from the stereoscopic display device may be used when rendering each pixel of the stereoscopic display device, e.g., to determine whether the pixel can be seen by the left eye position and to use the first frame, or whether the pixel can be seen by the right eye position and to use the second frame. Through these processes, the observer’s left eye is able to see the first video signal and the observer’s right eye is able to see the second video signal. The observer is able to benefit from stereopsis resulting from the difference in viewing angle to see a three-dimensional representation in real time of the optical outputs. The optical outputs may be left and right channels of a slit lamp microscope. The reading of the head tracking parameter from the stereoscopic display device may comprise using a routine (calling a routine) to obtain a parameter from the stereoscopic display device, for example, which might be a Spatial Reality Display that is available from the Sony Corporation and uses software marketed under the name “Unity” which obtains a “_WorldSpaceCameraPos” parameter (World Space Camera Position variable). A value of the parameter may be used to direct the selection of first frames and second frames during the rendering process. Viewed from a fifth aspect there is a stereoscopic microscope system. The system is for livestreaming an object on a stereoscopic display device to an observer. The system comprises a microscope, a stereoscopic display device and a computer. The microscope comprises a binocular viewer, a first image sensor for capturing a first livestream of the object, and a second image sensor for capturing a second livestream of the object. The stereoscopic display device is for displaying the stereoscopic livestream to the observer. The stereoscopic display device comprises a head tracking device for tracking the head position of the observer. The computer comprises a data processing device (such as a processor) and a memory. The memory comprises instructions that, when executed on the data processing device, cause the data processing device to initialise a computer model of the object and a texture mapping for rendering the computer model with a texture and repeatedly execute the following steps. The data processing device captures a first image of the object with the first image sensor, captures a second image of the object with the second image sensor, receives head tracking data of the observer from the head tracking device, determines one or more render camera viewpoints relative to the computer model based on the head tracking data, renders a stereoscopic image based on the computer model and the texture mapping using the one or more render camera viewpoints, wherein the left image of the stereoscopic image is rendered using the first image as the texture and the right image of the stereoscopic image is rendered using the second image as the texture, and displays the stereoscopic image on the stereoscopic display device to the observer as a livestream. The microscope may be a slit lamp microscope and the object may be an eye. The first image sensor and the second image sensor may be parts of a stereo camera arrangement mounted between a body of the microscope and the binocular viewer. The stereo camera arrangement may comprise a pair of beam splitters, each deflecting a portion of light in a left or right side of the microscope toward one of the image sensors. The stereo camera arrangement may comprise a bayonet fitting on one side for attachment to the body of the microscope and a bayonet fitting on an opposed side for attachment of the binocular viewer. The binocular viewer and the stereo camera may be arranged to capture the same field of view at the same magnification. The stereoscopic display device may comprise a pixel array comprising a plurality of pixels for displaying the stereoscopic image, and a micro optical lens array arranged over the pixel array and comprising a plurality of micro optical lenses, wherein each micro optical lens is arranged over multiple pixels of the pixel array and configured to project light from the multiple pixels over different viewing angles. The pixel array and micro optical lens array of the stereoscopic display device may provide a screen which is mounted at 45° to a horizontal surface. The head tracking sensor may be provided in a frame around the screen of the stereoscopic display device. The head tracking sensor may be mounted on the stereoscopic display device above the screen in a central location on the stereoscopic display device. According to a sixth aspect, there is provided a computer program product containing instructions that, when executed on a data processing device (such as a processor) of a computer, cause the data processing device to initialise a computer model of an object and initialise a texture mapping for rendering the computer model with a texture and repeatedly execute the following steps. The data processing device is caused to capture a first image of the object with a first image sensor, capture a second image of the object with a second image sensor, receive head tracking data of the observer from a head tracking device, determine one or more render camera viewpoints relative to the computer model based on the head tracking data, render a stereoscopic image based on the computer model and the texture mapping using the one or more render camera viewpoints, wherein the left image of the stereoscopic image is rendered using the first image as the texture and the right image of the stereoscopic image is rendered using the second image as the texture, and display the stereoscopic image on the stereoscopic display device to the observer as a livestream. According to a seventh aspect, there is provided a method of displaying a stereoscopic image from a microscope on a stereoscopic display device to an observer viewing a screen of the stereoscopic display device. The microscope comprises a left image sensor and a right image sensor, and the screen comprises a plurality of pixels arranged as an array and a lenticular lens comprising a plurality of lenses arranged over the pixels, each lens extending over at least two pixels in a horizontal direction of the screen and configured to project light from each of the at least two pixels over different regions horizontally. The method comprises: receiving head tracking data from a head tracking device on the stereoscopic display; determining from the head tracking data, for each pixel, whether the pixel can be seen from a left eye or a right eye of an observer based on which of the left eye position and right eye position is dominantly within the pixel’s projected region; and rendering each pixel, wherein: if the pixel can be seen by the left eye, then the pixel is rendered according to a left image from the left image sensor; or if the pixel can be seen by the right eye, then the pixel is rendered according to a right image from the right image sensor. Dominantly within the pixel’s projected region may refer to the eye position which is closest to the centre of the projected region. Dominantly within the pixel’s projected region may also refer to a predetermined eye of the left or right eye when both are within the projected region, for instance the right eye as this is typically the dominant eye of an observer. The microscope may be a slit beam microscope. The left and right image sensors may be mounted in a housing of a stereographic camera arranged between a body of the microscope and eyepieces of a binocular viewer, the stereographic camera comprising a pair of beam splitters that are arranged to intercept a portion (for example, between 25-75%, e.g., 50%) of the light being projected towards each of the eyepieces and direct the light towards the image sensors. The image sensors may be video image sensors arranged on opposite sides of the housing, the video feeds from which are coupled to a data processing device such as a processor of a computer. The computer may control the display of the stereographic images seen on the screen of the stereoscopic display device based on the head tracking data output from the head tracking device on the stereoscopic display device. The different regions that the pixels may project light over as a result of the overlying lens may project in angular directions that are at least 3° apart, for example more than 5° apart, in the horizontal direction of the screen. There may be four or more pixels arranged under each lens in the horizontal direction across a row of pixels and the method may include rendering each of the pixels using the left image or the right image as basis for the rendering, software taking into account deflection of the light from the pixel as it passes through material of the lens and towards the observer’s eye to collectively build up a left image I right image in the eye of the observer (interlacing slices of the image). Each pixel may be made of sub-pixels, for example, comprising red, green and blue sub-pixels. The computer may render the stereographic images automatically in real-time (without noticeable lag, e.g., less than 0.5s) so that the observer can use the stereoscopic display as a live display of the object that is being viewed by the microscope, for example, the object might be an eye that is being examined. The observer may be able to view a three-dimensional representation of the object on the stereoscopic display device as a result of the head tracking data being used to select either the left image from the left image sensor or the right image from the right image sensor as basis for rendering each pixel on the screen of the stereoscopic display device. The left and right images being projected to the left and right eyes of the observer may then allow the observer to see a three-dimensional representation by means of stereopsis. Depth effects may be enhanced using a computer model. According to an eighth aspect there is provided a stereoscopic microscope system for putting the method of the seventh aspect into effect. The stereoscopic microscope system may comprise a microscope, a stereoscopic display device and a computer. The microscope may comprise a binocular viewer, a first image sensor for capturing a first livestream of the object, and a second image sensor for capturing a second livestream of the object. The stereoscopic display device is for displaying the stereoscopic livestream to the observer. The stereoscopic display device may comprise a head tracking device for tracking the head position of the observer. The stereoscopic display device may comprise a screen with a pixel array comprising a plurality of pixels for displaying the stereoscopic image, a micro optical lens array arranged over the pixel array and comprising a plurality of micro optical lenses, wherein each micro optical lens is arranged over multiple pixels of the pixel array and configured to project light from the multiple pixels over different viewing angles in a horizontal direction of the screen. The computer may comprise a data processing device and a memory, the memory comprising instructions that, when executed on the data processing device, causes the data processing device to perform the method steps of the seventh aspect, namely: receiving head tracking data from a head tracking device on the stereoscopic display; determining from the head tracking data, for each pixel, whether the pixel can be seen from a left eye or a right eye of an observer based on which of the left eye position and right eye position is dominantly within the pixel’s projected region; and rendering each pixel, wherein: if the pixel can be seen by the left eye, then the pixel is rendered according to a left image from the left image sensor; or if the pixel can be seen by the right eye, then the pixel is rendered according to a right image from the right image sensor. The stereoscopic microscope may be a slit beam microscope and the stereo camera arrangement may be as described above, comprising a pair of beam splitters, each for deflecting light toward one of the image sensors, the first and second image sensors being arranged opposite each other on left and right sides of the stereo camera arrangement. The stereoscopic display device may be a full colour lenticular stereoscopic display device as described above which may be arranged with the screen inclined at 45° to a horizontal / vertical plane. Features described herein with respect to one of the aspects may equally apply to the other aspects, and such combinations are thus contemplated herein (even if not explicitly recited in combination). BRIEF DESCRIPTION OF THE DRAWINGS A detailed description of one or more embodiments are presented herein by way of exemplification and not limitation with reference to the Figures. The following descriptions are given by way of example and should not be considered limiting in any way: Fig. 1 shows an example slit lamp microscope having a stereoscopic camera; Fig. 2 shows an exemplary image of an eye; Fig. 3 shows a schematic of the stereoscopic microscope system having a stereoscopic display device; Fig. 4 shows a flowchart of a method for rendering a stereoscopic image for displaying on a stereoscopic display device; Fig. 5 shows a representation of head tracking of an observer; Fig. 6 shows a perspective view of the head tracking of the observer in Fig. 5; Fig. 7 shows an exemplary render scene; Fig. 8 shows a flowchart of a method for rendering a stereoscopic livestream for displaying on a stereoscopic display device; Figs. 9A and 9B show a stereoscopic image of a scene; and Fig. 10 shows a schematic representation of a lenticular stereoscopic display device. For ease of reference, like reference numerals are used to indicate the same or substantially similar features in each of the figures. DETAILED DESCRIPTION Figure 1 illustrates a slit lamp microscope 100 with a stereoscopic camera 10 (e.g., stereo camera system or imaging system) installed between a microscope body 101 and a binocular eyepiece 102 of the slit lamp microscope 100. In this example, the stereoscopic camera 10 is arranged to intercept the stereoscopic beams of light (e.g., two separate optical paths from respective left and right objectives output from the microscope body 101) by splitting the beams of light before the light reaches the binocular eyepiece 102. The slit lamp microscope 100 is used in many applications, particularly in the observing of an eye 200 of a patient. Figure 2 illustrates an exemplary image of an eye as seen through a slit lamp microscope 100 when used to observe the eye 200. In one example, the stereoscopic camera 10 is provided in a stereoscopic microscope system 300 as shown in Figure 3. The stereoscopic camera 10 includes a housing 12 (e.g., a self-contained housing containing within itself all parts necessary for imaging). A first image sensor 14a and a second image sensor 14b are mounted within the housing 12. The stereoscopic camera 10 includes a first beam splitter 16a and a second beam splitter 16b, one each for the respective first image sensor 14a and second image sensor 14b. Each beam splitter 16a, 16b may comprise two triangular prisms and a reflective coating or layer positioned at an interface between the two triangular prisms. Each triangular prism may be optically transparent. The reflective coating or layer may be configured to reflect a portion of incident light. The reflective coating or layer may be configured to transmit a portion (another portion) of incident light. The two triangular prisms may together form a substantially cuboid or cube shape. That is, the beam splitter 16a, 16b may be substantially cuboid or cube shape. An optical path to the respective image sensor may be provided through the substantially cuboid or cube shape. The optical path may comprise an incident beam of light and a reflected beam of light (a portion of the incident beam of light being reflected) together forming an optical plane (e.g., a two-dimensional plane). The optical path to the respective image sensor 14a, 14b may follow an L-shaped path. A portion of the incident beam of light may also pass straight through to an eyepiece 102 of a binocular viewer. Figure 3 illustrates an exemplary stereoscopic microscope system 300 that comprises the slit lamp microscope 100, the stereoscopic camera 10, a computer 310, a stereoscopic display device 330 viewed by an observer 360 for displaying a stereoscopic image 340 of the eye 200, and a head tracking device 350 for tracking the head position of the observer 360. The computer 310 comprises a data processing device such as a processor 312, a memory 314, a storage 316, and an input / output interface (I / O) 318. The computer is connected, via the I / O 308, to the first and second image sensors 14 of the stereoscopic camera 10, to the stereoscopic display device 330, and the head tracking device 350. In practice, the computer 310 (shown schematically in the figure) may well comprise further parts, and the input / output interface may be supplemented, at least in part, by a wireless data transmitter / receiver system, such as a Bluetooth connection, e.g., to transmit data between components of the system, such as the image sensors 14a, 14b, the database 320, the stereoscopic display device 330 and the head tracking device 350. The stereoscopic display device 330 may comprise the head tracking device 350 as shown. The head tracking device 350 may be integrated into a housing 332 of the stereoscopic display device 330. It may be positioned centrally above the screen 334 of the stereoscopic display device 330, i.e., midway in a horizontal direction and in a border 336 of the housing 332 above the screen 334. The screen may be mounted at 45° within a stand 338 so that the observer 360 can see the three-dimensional visualisation easily. The stereoscopic display device 330 displays different images to the left 364 and right 366 eyes of the observer 360, such that a three-dimensional visualisation is produced by means of stereopsis. The example stereoscopic display device 330 in this instance is a screen utilising lenticular lens comprising a plurality of lenses, such as a Spatial Reality Display produced by Sony. Other stereoscopic display devices such as a virtual reality headset are also contemplated herein where wearing a headset is not a complication for the observer. However, a lenticular stereoscopic display device 330, such as the one shown in Figure 3, the functionality of which is illustrated in Figure 10, has an advantage that the stereoscopic image 340 can be seen without the use of polarised glasses or a specialised headset. The head tracking allows the stereoscopic image to be seen easily from a wide range of viewing angles rather than only when still and central to the display. As such, a lenticular stereoscopic display device 330 provides a simplified way to illustrate a stereoscopic image to an observer 360, who might be a student learning how to examine a patient’s eye using the slit lamp microscope. The observer 360 may be an ophthalmologist using both stereoscopic images to help their examination of the patient’s eye 200. The observer 360 can switch easily from observations on the stereoscopic display device 330 to observations using the binocular eyepiece 102 without having to remove I put on glasses or a special head set. For example, the stereoscopic display device 330 can be used to identify features of interest in the eye 200 and then the observer can use the binocular eyepiece 102 to examine such a feature of interest in more detail, switching easily between the three-dimensional electronic stereoscopic image and the optical stereoscopic image by moving their head from one to the other without having to put on I remove glasses or a head set. A lenticular stereoscopic display device 330 such as that illustrated in Figure 3 also takes up only a small area of a table or counter surface supporting the stereoscopic microscope system 300 and is a relatively cheap option for the ophthalmologist. Other forms of stereoscopic display device are known, for example, stereoscopic projector systems, but these tend to be more expensive and take up much more room on the table or counter surface. They also require the observer to view the stereoscopic image from a relatively precise position to be able to observe the stereoscopic effect. The computer 310 receives first and second images from the first and second image sensors 14a, 14b, and head tracking data from the head tracking device 350. These may be received in a livestream manner. By livestream manner, it is meant that the first and second images are received “live”, i.e., without a noticeable time lag that would affect the viewing experience of the observer 360. The first and second images may be time-synced frames of first and second video streams received from the first and second image sensors 14a, 14b. Time-synced frames should be understood as occurring roughly at the same point in time such that any temporal differences are not noticeable to a human observer 360. This can include one frame occurring immediately after the other, for example, through the software calling one image sensor 14 for a frame and then the other 14 which will be treated as time-synced frames. The rate of frame sampling may be 50 Hz or more. The time-synced frames may be saved to storage device 316 of the computer 310 and / or database 320 for later play back, for example, when reviewing a previous live examination of a patient’s eye conducted on the stereoscopic microscope system 300 at a later time, or shared from another ophthalmologist using another stereoscopic microscope system either as a livestream or a pre-recorded video stream. Based on the received first and second images, and the head tracking data, the computer 310 may render the stereoscopic image 340 in accordance with the method 400 of Figure 4. The stereoscopic image 340 may subsequently be displayed on the on the stereoscopic display device 330 to the observer 360. At the same time, the stereoscopic image 340, for example, as a pair of time-synced frames for the observer’s 360 left and right eyes, may also be saved in the storage 316 of the computer 310 and / or a database 320 connected via a network to the I / O 308 of the computer. When the first and second images, and head tracking data are received in a livestream manner, the stereoscopic image 340 is repeatedly rendered such that a stereoscopic video of the object 200 is generated. Figure 4 illustrates an exemplary method 400 of rendering the stereoscopic image 340 of the eye 200 for displaying on the stereoscopic display device 330 to the observer 360 viewing the stereoscopic display device 330. The stereoscopic image 340 comprises a left image and a right image. The method 400 may comprise: 410 initialising a computer model 720 of the object 200 (e.g. the eye 200) and a texture mapping for rendering the computer model with a texture; 420 receiving a first image of the object 200 from a first image sensor (e.g. the first image sensor 14a of the stereoscopic camera 10 of the stereoscopic microscope system 300); 430 receiving a second image of the object 200 from a second image sensor (e.g. the second image sensor 14b of the stereoscopic camera 10 of the stereoscopic microscope system 300); 440 receiving head tracking data of the observer (e.g. from the head tracking device 350 of the stereoscopic display device 330); 450 determining one or more render camera viewpoints 740, 750 relative to the computer model 720 based on the head tracking data; and 460 rendering the stereoscopic image based on the computer model 720 and the texture mapping using the one or more render camera viewpoints 740, 750. The left image of the stereoscopic image is rendered using the first image as the texture and the right image of the stereoscopic image is rendered using the second image as the texture. The method may involve repeating the steps 420-460, using the same computed model initialised in step 410. Figure 5 schematically illustrates the received head tracking data, with Figure 6 being an aerial perspective of this. In the head tracking data, a global coordinate frame 352 may be defined relative to the head tracking device 350 and thus relative to the stereoscopic display device 330. This global coordinate frame 352 is a fixed point in space, which may correspond to the stereoscopic display device 330 when this is a stationary display or may correspond to an arbitrary fixed point in space when the stereoscopic display device 330 movable, e.g. when it is a virtual reality headset. Within this global coordinate frame 352, the head tracking device 350 monitors the observer 360 to determine an observer coordinate frame 362 indicative of a position of the observer 360, and a pair of eye positions 364, 366. Based on the observer coordinate frame 362, a sagittal plane 365 of the observer may be determined, which evenly divides the head of the observer into right and left sections. The sagittal plane 365 of the observer 360 may be used as a separation plane. The pair of eye positions 364, 366 are compared to the sagittal plane 365 and, based upon which side of the sagittal plane 365 that the eye positions lie, a left eye position 364 and a right eye position 366 is determined. The observer coordinate plane 362 may be a matrix transformation using homogenous coordinates, and the left 364 and right 366 right eye positions may be a coordinate vector. The sagittal plane may be represented by a plane equation in the observer coordinate plane 362. Figure 7 illustrates an exemplary render scene 700 based on the head tracking data which is rendered in the method 400 to produce the stereoscopic image 340. In this render scene there is a render coordinate frame 710, a computer model 720 with a UV texture mapping, an image plane 730, a left render camera viewpoint 740, and a right camera render viewpoint 750. The render coordinate frame 710 corresponds to the global coordinate frame 352, in which the position and pose of the computer model 720 may be fixed. The stereoscopic image is rendered on the image plane 730, the position of which corresponds to the position of the stereoscopic display device 330 used to display the stereoscopic image 340. For example, the image plane 730 corresponds to the pixel array 334 and micro optical lens array 336 as discussed in relation to the illustration of the stereoscopic display device 330 in Figure 10. However, where a virtual reality headset is used, the position of the image plane 730 will correspond to that of the virtual reality headset. The left 740 and right 750 render camera viewpoints are determined based on the head tracking data, and their positions correspond to the left eye 364 and the right eye 366 of the observer 360. The left render camera viewpoint 740 is used to render the left image of the stereoscopic image based on the computer model 720, wherein the first image is used as the texture for the computer model 720 with the UV texture mapping. The right render camera viewpoint 750 is used to render the right image of the stereoscopic image based on the computer model 720, wherein the second image is used as the texture for the computer model 720 with the UV texture mapping. A render camera viewpoint is a position in the render coordinate frame 710 which determines how the computer model 720 is projected onto the image plane 730, as illustrated by the dashed lines in Figure 7. The rendered stereoscopic image 340, when displayed on the stereoscopic display device 330, therefore provides a three-dimensional visualisation of the eye 200. When the position of the computer model 620 is fixed relative to the render coordinate frame 710 and thus the global coordinate frame 352, this has the effect of fixing the position of the resulting three-dimensional visualisation to a point in physical space. Thus, when rendering the computer model 620 with the position of image plane 730 corresponding to the position of the stereoscopic display device 330, and the left 740 and right 750 render camera viewpoints correspond to the positions of the left eye 364 and the right eye 366 of the observer; the three dimensional visualisation appears at this point in physical space, seen through the stereoscopic display device 330. The three-dimensional visualisation is updated for the new head position, the observer benefitting from stereopsis to view the three-dimensional visualisation from the updated position. Depending on processing power available, the model may be updated and moved with respect to the observer in virtual space. Thus, as the observer 360 moves around and the rendered stereoscopic image 340 is updated with live head tracking data, the observer may be able to move around the three-dimensional visualisation as if it were an actual object at that point in space. Figure 8 illustrates an exemplary method 800 of rendering a stereoscopic livestream of the eye 200 for displaying on the stereoscopic display device 330 to the observer 360 viewing the stereoscopic display device 330. The method 800 may comprise: 810 initialising a computer model 720 of the object 200 (e.g. the eye 200) and a texture mapping for rendering the computer model with a texture; 820 capturing a first image of the object 200 with the first image sensor (e.g. the first image sensor 14a of the stereoscopic camera 10 of the stereoscopic microscope system 300); 830 capturing a second image of the object 200 with the second image sensor (e.g. the second image sensor 14b of the stereoscopic camera 10 of the stereoscopic microscope system 300); 830 receiving head tracking data of the observer from a head tracking device (e.g. the head tracking device 350 of the stereoscopic display device 330); 840 determining one or more render camera viewpoints 740, 750 relative to the computer model 720 based on the head tracking data; 850 rendering a stereoscopic image based on the computer model 720 and the texture mapping using the one or more render camera viewpoints 740, 750, wherein the left image of the stereoscopic image is rendered using the first image as the texture and the right image of the stereoscopic image is rendered using the second image as the texture; and 860 displaying the stereoscopic image on the stereoscopic display 330 device to the observer. The steps 820-860 are repeatedly executed such that the displayed stereoscopic image is updated in real-time such that the stereoscopic image is displayed on the stereoscopic display 330 device to the observer as a livestream. To illustrate schematically how a stereoscopic display device 330 can display a stereoscopic image, Figures 9A and 9B show a stereoscopic image of a scene. In particular, Figure 9A schematically depicts a chess scene 900 and the position of a left eye 364 and a right eye 366, with Figure 9B depicting how this chess scene 900 appears to the left eye 364 and the right eye 366, i.e. a stereoscopic image of the chess scene 900 with a left image 910 and a right image 920. Figure 10 shows a schematic representation of a lenticular stereoscopic display device 1000 comprising a pixel array 1100 and a lenticular lens 1200 comprising a plurality of lenses, each lens being arranged over a plurality of pixels. When the lenticular stereoscopic display device 1000 is viewed from a particular viewpoint 1300, only certain pixels behind each lens are visible due to the manner in which the lenses project (i.e. refract) the light emitted from each pixel. This enables the lenticular stereoscopic display device 1000 to simultaneously display different images 1400 at different viewpoints 1300 by slicing each image 1400 vertically and displaying them in an interlaced manner on the pixel array 1400. Although this disclosure has been described in terms of preferred examples, it should be understood that these examples are illustrative only and modifications and alterations are possible within the scope of the claims.

Claims

1. A method of rendering a stereoscopic image of an object for displaying on a stereoscopic display device to an observer viewing the stereoscopic display device, the stereoscopic image comprising a left image and a right image, the method comprising, on a data processing device, initialising a computer model of the object and a texture mapping for rendering the computer model with a texture, the method comprising:receiving a first image of the object from a first image sensor;receiving a second image of the object from a second image sensor; receiving head tracking data of the observer;determining one or more render camera viewpoints relative to the computer model based on the head tracking data; andrendering the stereoscopic image based on the computer model and the texture mapping using the one or more render camera viewpoints, wherein the left image of the stereoscopic image is rendered using the first image as the texture and the right image of the stereoscopic image is rendered using the second image as the texture.

2. A method according to claim 1, wherein the method is performed during an eye examination, the first and second image sensors are part of a slit lamp microscope comprising a stereo camera arrangement, the object is an eye, and the stereo camera arrangement houses the first image sensor spaced apart from the second image sensor.

3. A method according to claim 2, wherein:the slit lamp microscope comprises a binocular viewer to view the object; and the stereo camera arrangement is mounted between a microscope body and the binocular viewer of the slit lamp microscope, the stereo camera arrangement comprising a pair of beam splitters, one for each optical channel of the slit lamp microscope, the method comprising diverting a portion of light from the optical channels away from the binocular viewer and towards the first and second image sensors, respectively.

4. A method according to claim 3, wherein the binocular viewer and the stereo camera are arranged to capture the same field of view at the same magnification.

5. A method according to any preceding claim, wherein the head tracking data of the observer tracks a position of the observer’s head relative to the stereoscopic display device.

6. A method according to any preceding claim, wherein the computer model of the object is a pre-determined computer model.

7. A method according to claim 6, wherein the pre-determined computer model is a bounded plane.

8. A method according to any preceding claim, wherein the head tracking data of the observer comprises a pair of eye positions of the observer and the one or more render camera viewpoints is a pair of render camera viewpoints, the method comprising:rendering the left image of the stereoscopic image based on a left render camera viewpoint of the pair of render camera viewpoints; andrendering the right image of the stereoscopic image based on a right render camera viewpoint of the pair of render camera viewpoints.

9. A method according to claim 8, the method comprising:determining a separation plane based on the head tracking data of the observer, wherein the separation plane corresponds to a sagittal plane of a head of the observer; anddetermining which of the pair of render camera viewpoints correspond to the left render camera viewpoint and the right render camera viewpoint based on which side of the separation plane each of the pair of render camera viewpoints lies.

10. A method of rendering a stereoscopic video for displaying on a stereoscopic display device to an observer viewing the stereoscopic display device, wherein each stereoscopic image of the stereoscopic video is rendered according to the method of any preceding claim.

11. A method according to claim 10, wherein the head tracking data is in real-time and the stereoscopic video is a livestream for displaying on the stereoscopic display device to the observer in real-time.

12. A method according to any preceding claim, the method comprising displaying the stereoscopic image or stereoscopic video on the stereoscopic display device to the observer.

13. A method according to claim 12, wherein the stereoscopic display device comprises:a pixel array comprising a plurality of pixels for displaying the stereoscopic image;a micro optical lens array arranged over the pixel array and comprising a plurality of micro optical lenses, wherein each micro optical lens is arranged over multiple pixels of the pixel array and configured to project light from the multiple pixels over different viewing angles; andthe head tracking data of the observer defines a pair of eye positions of the observer, wherein the pair of eye positions is a left eye position and a right eye position;wherein the method comprises:interlacing the left image and the right image of the stereoscopic image;displaying the interlaced stereoscopic image on the pixel array; andprojecting, with the micro optical lens array, the left image towards the left eye position, and the right image towards the right eye position.

14. A method of rendering a stereoscopic image of an object for displaying on a stereoscopic display device to an observer viewing the stereoscopic display device, the stereoscopic image comprising a left image and a right image, the method comprising:receiving a first image of the object from a first image sensor;receiving a second image of the object from a second image sensor, wherein the first image sensor is spaced apart from the second image sensor in a stereo camera arrangement;receiving head tracking data of the observer from the stereoscopic display device, wherein the head tracking data describes a pair of eye positions of the observer;determining a pair of render camera viewpoints relative to a computer model based on the pair of eye positions;rendering a pixel of the stereoscopic image based on the computer model and a texture mapping, the rendering using the head tracking data to select the texture to apply with the texture mapping, wherein:if the pixel is determined to be of the left image of the stereoscopic image, the pixel is rendered using the first image as the texture and a left camera viewpoint; andif the pixel is determined to be of the right image of the stereoscopic image, the pixel is rendered using the second image as the texture and a right camera viewpoint.

15. A method according to claim 14, the method comprising:determining a separation plane based on the head tracking data of the observer, wherein the separation plane corresponds to a sagittal plane of a head of the observer; anddetermining which of the pair of render camera viewpoints correspond to the left render camera viewpoint and the right render camera viewpoint based on which side of the separation plane each of the pair of render camera viewpoints lies.

16. A method of generating a video feed for a stereoscopic display device, the video feed being for displaying a three-dimensional representation of an object to an observer viewing a screen of the stereoscopic display device, the method comprising:receiving a first video signal from a first image sensor positioned to observe a first optical output of an object taken from a first channel of a stereoscopic microscope where the object is being viewed at a first viewing angle, the first video signal comprising a sequence of first frames;receiving a second video signal from a second image sensor positioned to observe a second optical output of the object from a second channel of a stereoscopic microscope where the object is being viewed at a second viewing angledifferent to the first, the second video signal comprising a sequence of second frames;processing first frames and corresponding second frames as frame pairs;reading a head tracking parameter from the stereoscopic display device that is indicative of left and right eye position for the observer, the parameter being output in real-time from the stereoscopic display device and read for at least each frame pair; andgenerating a video feed of stereoscopic image frames for the screen,wherein the screen comprises a lenticular display of lenses arranged over an array of pixels, the lenses being configured to display, in use, a first image to a left eye of the observer and a second image to a right eye of the observer, andwherein the head tracking parameter from the stereoscopic display device is used in a rendering process to automatically select either the first frame or the second frame of each frame pair as basis when rendering each pixel of the display.

17. A system for livestreaming an object on a stereoscopic display device to an observer, the system comprising:a microscope comprising: a binocular viewer, a first image sensor for capturing a first livestream of the object, and a second image sensor for capturing a second livestream of the object;a stereoscopic display device for displaying the stereoscopic livestream to the observer, the stereoscopic display device comprising a head tracking device for tracking the head position of the observer; anda computer comprising a data processing device and a memory, the memory comprising instructions that, when executed on the data processing device, causes the data processing device to initialise a computer model of the object and a texture mapping for rendering the computer model with a texture and repeatedly execute the following steps:capture a first image of the object with the first image sensor;capture a second image of the object with the second image sensor;receive head tracking data of the observer from the head tracking device;determine one or more render camera viewpoints relative to the computer model based on the head tracking data;render a stereoscopic image based on the computer model and the texture mapping using the one or more render camera viewpoints, wherein the left image of the stereoscopic image is rendered using the first image as the texture and the right image of the stereoscopic image is rendered using the second image as the texture; anddisplay the stereoscopic image on the stereoscopic display device to the observer as a livestream.

18. A system as claimed in claim 17, wherein the microscope is a slit lamp microscope and the object is an eye.

19. A system as claimed in claim 17 or 18, wherein the first image sensor and the second image sensor are parts of a stereo camera arrangement mounted between a body of the microscope and the binocular viewer.

20. A system as claimed in claim 19, wherein the stereo camera arrangement comprises a pair of beam splitters, each for deflecting light toward one of the image sensors, the first and second image sensors being arranged opposite each other on left and right sides of the stereo camera arrangement.

21. A system as claimed in claim 19 or 20, wherein the binocular viewer and the stereo camera arrangement are arranged to capture the same field of view at the same magnification.

22. A system as claimed in any of claims 19 to 21, wherein the stereoscopic display device comprises:a pixel array comprising a plurality of pixels for displaying the stereoscopic image, anda micro optical lens array arranged over the pixel array and comprising a plurality of micro optical lenses,wherein each micro optical lens is arranged over multiple pixels of the pixel array and configured to project light from the multiple pixels over different viewing angles.

23. A system as claimed in claim 22, wherein the micro optical lens array is an array of lenticular lenses.

24. A system as claimed in claim 23, wherein the pixel array and micro optical5 lens array of the stereoscopic display device provide a screen which is mounted at 45° to a horizontal surface.

25. A system as claimed in claim 24, wherein the head tracking sensor is provided in a frame around the screen of the stereoscopic display device, optionally io wherein the head tracking sensor is mounted on the stereoscopic display deviceabove the screen in a central location on the stereoscopic display device.34

Images