A method and system for processing image data

Abstract

A computer-implemented method and system for processing image data in optical coherence tomography. An initial image data captured by employing an optical coherence tomography based imaging technique is received. The image data is modified using a modification function that is configured to change magnitude while maintaining phase of the image data, in order to obtain a modified image data, wherein the modification function is selected based on one or more image quality metrics; applying a filtering process to the modified image data using a filter so as to obtain a filtered image data. A compensation is performed on the filtered image data in order to compensate for alterations resulting from the initial modification of the image data prior to filtering, wherein the compensation is performed based on the modification function and the filter.

Description

[0001]P136670PC00 Title: a method and system for processing image data FIELD OF THE INVENTION The invention relates to a computer implemented method and system for processing image data. The image data may be obtained by performing optical coherence tomography for instance. The invention also relates to an imaging device and a computer program product. BACKGROUND TO THE INVENTION In the field of optical imaging, such as for instance optical coherence tomography (OCT), achieving high-quality images is a critical yet challenging endeavor. This challenge becomes more pronounced in Fourier-domain OCT techniques, where imaging artifacts and unwanted background signals are commonplace. These issues may stem from parasitic reflections within the imaging setup or probe, which are often unavoidable and significantly degrade image quality. The point-spread function (PSF), an essential aspect of imaging quality, is dictated by the spectral shape in OCT systems. Traditional methods, such as windowing and de-convolution, are employed in spectrometer-based systems to mitigate side-lobes and artifacts and to refine the PSF. However, these techniques are not universally applicable or result in a degradation of resolution. Swept- source, point-scanning, line-field, and full-field OCT systems may lack access to the precise spectral shape necessary for these operations. The absence of this information leads to further degradation of the image quality in these systems. Moreover, some of these techniques lack a (full) confocal gating which may result in reflections on the camera, which vary with lateral position, exacerbating the problem of image artifacts. These imaging artifacts not only compromise the quality of the images but also influence the cost-efficiency of OCT systems. Reducing artifacts typically necessitates the use of more expensive components or may require a thorough and manual selection and matching of components, elevating the overall costs. To address these challenges, several methods have been explored. These include periodic acquisition of the reference spectrum, subtracting median or average signals, optimizing the spectral shape either electronically, optically or digitally, and spatial filtering. Each of these methods, while offering some relief, comes with its own set of limitations and complexities. For example, in the context of swept-source OCT, the ability to shape the spectrum can be achieved by controlling the current through the Semiconductor Optical Amplifier (SOA). By adjusting the current, it is possible to finely tune the spectral shape, which is important for optimizing the point-spread function (PSF) and, consequently, the image quality. In contrast, in Spectral-Domain (SD) OCT systems, changes in the current also affect the spectrum, but this process is somewhat less controlled. The uncontrolled variation in the spectral shape in SD systems can lead to challenges in maintaining consistent image quality, as the PSF is directly influenced by these spectral changes. A particularly important aspect in many OCT applications is the preservation of phase information. However, some of the existing artifact reduction methods focus solely on intensity data, neglecting the phase aspect. Such intensity- based methods can inadvertently introduce negative image artifacts, further complicating the image processing landscape. Given these challenges, there is a strong desire for improved methods that can effectively tackle both background and artifact removal, as well as axial PSF shaping in image date, for example obtained via OCT imaging. The current state of the art, despite its advancements, still falls short in offering a comprehensive solution that addresses these issues across various imaging systems, including for example full-field Fourier-domain systems, especially with multimode fiber illumination, and commercial swept-source OCT systems. It is needed that the method can effectively ensure enhanced image quality, cost- effectiveness, and applicability across a broad range of imaging systems. SUMMARY OF THE INVENTION It is an object of the invention to provide for a method and a system that obviates at least one of the above mentioned drawbacks. Additionally or alternatively, it is an object of the invention to provide for an improved computer implemented method and system for processing image data, for example in optical coherence tomography. Additionally or alternatively, it is an object of the invention to provide for an improved computer implemented method and system that can more effectively filter, whilst preventing or reducing the creation of major side-lobes. Additionally or alternatively, it is an object of the invention to provide for an improved computer implemented method and system that can apodize (cf. remove side-lobes) without sacrificing resolution. Thereto, the invention provides for a computer-implemented method for processing image data in optical coherence tomography, comprising: receiving an initial image data captured by employing an optical coherence tomography based imaging technique; modifying the image data using a modification function that is configured to change magnitude while maintaining phase of the image data, in order to obtain a modified image data, wherein the modification function is selected based on one or more image processing metrics; applying a filtering process to the modified image data using a filter so as to obtain a filtered image data; and performing a compensation on the filtered image data in order to compensate for alterations resulting from the initial modification of the image data prior to filtering, wherein the compensation is performed based on the modification function and the filter. Initial OCT-based image data is received, then a modification function is applied to alter the magnitude while preserving the phase of this data, thereby generating modified image data. The selection of the modification function is guided by specific image processing / quality metrics. Subsequently, a filtering process is applied to the modified image, followed by a compensation step to rectify alterations introduced during the initial modification. This comprehensive process facilitates enhanced image clarity and quality in OCT imaging. This can enhance image quality in optical coherence tomography (OCT) through improved background subtraction and point-spread function (PSF) optimization for example. The process begins with the acquisition of initial image data using an OCT-based imaging technique. This data is typically complex-valued, reflecting both amplitude and phase information inherent to OCT imaging. The next step involves modifying this initial image data. The modification is executed by a specially designed function that alters the magnitude of the image data while maintaining its phase. The preservation of the phase is important as it contains important information about the structure and properties of the imaged sample that are used in various OCT-based techniques, such as Doppler OCT, polarization- sensitive OCT, or phase-sensitive optoretinography. Altering the magnitude while keeping the phase intact allows for the enhancement of certain image features (like contrast or clarity) without losing integral spatial and depth information. The choice of the modification function is not arbitrary. It is carefully selected based on specific image processing / quality metrics. This selection can be important because different imaging scenarios may require different types of enhancement. For example, in some cases, sparsity (L1 metric) might be the desired metric, focusing on enhancing the most significant features of the image while suppressing less important details. In other cases, metrics like Shannon entropy or a power function might be more appropriate, depending on the characteristics of the initial image data. By tailoring the modification function to these metrics, the method ensures that the image enhancement process is precisely aligned with the specific needs and characteristics of the image data. This targeted approach allows for a more effective and efficient enhancement process, leading to higher quality results. Following the modification of the image data, a filtering process is applied. This step is designed to further refine the image, such as by removing unwanted noise or artifacts, thereby enhancing the overall quality of the image. The final stage involves performing compensation on the filtered image data. Advantageously, this step addresses and corrects any alterations or distortions introduced by the initial modification of the image data. By compensating for these alterations, the method can effectively ensure that the final image retains the integrity and accuracy of the original data, while benefiting from the enhancements introduced through the modification and filtering processes. The method provides for a comprehensive and adaptable approach to image processing in OCT, delivering enhanced image quality through careful and precise manipulation of the image data at various stages. Advantageously, high precision and consistency in image processing can be obtained, making it a beneficial tool in the field of at least OCT imaging. In some examples, the modification function can vary based on different imaging scenarios, such as varying the degree of magnitude alteration for different tissue types in medical imaging or different materials in industrial applications. It will be appreciated that the term “weights” and the concept of local "weighting" may also be used for describing the modification process. By conceptualizing the modification function as “weights,” it becomes clear that the initial image data is being altered through a process of assigning varying levels of importance or emphasis to different parts of the data. This local “weighting” implies a selective enhancement or suppression of specific image features, which is achieved by multiplying the original data with these weights. The use of “weighting” underscores that the modification is not a uniform alteration of the entire image but a nuanced, area-specific adjustment. This interpretation aligns with the subsequent steps of the method, where division in the compensation stage serves as a counteraction to the multiplicative effects of the initial “weighting.” Optionally, performing the compensation includes: determining a correction derived from applying a process analogous to the filtering process to the modification function; and adjusting the filtered image data using this correction to address alterations caused by the initial modification of the image data. A correction factor can be derived by applying a process analogous to the filtering process, but to the modification function itself. This correction can then be utilized to adjust the filtered image data, thereby addressing any distortions or alterations caused by the initial modification of the image data. This step can ensure the fidelity and accuracy of the final image after processing. Advantageous, precise compensation for alterations caused by initial data modification can be performed. This precision is achieved by applying a process analogous to the filtering process to the modification function itself and using the resultant correction to adjust the filtered image data. This step ensures that any distortions introduced during the initial modification are accurately counterbalanced, thereby enhancing the fidelity of the processed image. This provides for the ability to mirror the effects of the primary filtering process onto the modification function. By doing so, it generates a correction that is inherently aligned with the alterations caused by the initial modification of the image data. This alignment ensures that the corrective adjustments are not only precise but also specifically tailored to counterbalance the distortions introduced during the initial modification phase. Consequently, when this tailored correction is applied to the filtered image data, it effectively neutralizes or reverses the alterations induced by the initial modification. This neutralization can be important as it restores the integrity of the image data, ensuring that any enhancements or alterations made during the filtering process reflect true improvements rather than artifacts of the initial modification. The result is an enhanced fidelity of the processed image, with distortions accurately counteracted, leading to a more accurate and reliable representation of the original image content. Optionally, the compensation step involves a calculated function to correct for any detrimental effects induced by the initial modification and subsequent filtering. Optionally, performing the compensation includes: applying the same filter used in the filtering process to the modification function to obtain a filtered reference; and dividing the filtered image data by a term indicative of and / or based on the filtered reference in order to compensate for alterations introduced by the initial modification of the image data. The initial modification of the image data may be performed through multiplication. The use of division for compensation can follow from the use of multiplication in the initial modification phase. In mathematical operations, division is often the inverse or corrective operation to multiplication. Therefore, when the modification of the image data initially involves multiplying certain values, it can follow that the compensation step would involve division. This division is used to counterbalance any distortions or alterations introduced by the initial multiplication, thereby restoring the integrity and accuracy of the image data. It will be appreciated that initial modification process can be described as “applying” a modification function, which encompasses a range of potential mathematical operations, including multiplication, which may be implied by the subsequent use of division for compensation. Compensation can involve using the same filter, as applied in the filtering process, to the modification function to generate a filtered reference. The filtered image data can then be divided by a term that is indicative of or based on this filtered reference assuming the original modification was performed by multiplying with the modification function. This division compensates for alterations introduced by the initial modification, ensuring an improved representation in the processed image. This provides for an effective normalization of filtered image data. The same filter can be used in the filtering process to the modification function to obtain a reference. The filtered image data is then divided by a term based on this reference, effectively normalizing the data and compensating for initial modifications. This process ensures a consistent and balanced approach to data alteration and correction. The filter used in the primary filtering process of the image data is also applied to the modification function. The modification function, as mentioned earlier, alters the magnitude of the image data while preserving its phase. By applying the same filter to this modification function, a reference is created. This reference represents how the modification function itself is affected by the filtering process. This creates a standard of comparison or a baseline, which reflects the extent to which the modification function is influenced by the filtering. It’s an important step to understand and quantify the specific changes the filtering process imposes on the modified image data. Dividing the filtered image data by a term derived from the filtered reference, normalizes the filtered image data against the changes induced by the initial modification. This division effectively compensates for any alterations or distortions introduced during the initial modification step. Since the filtered reference encapsulates the effect of the filter on the modification function, dividing by this reference importantly ‘undoes’ these effects in the filtered image data, ensuring that the final image accurately represents the original scene or structure being imaged. The process of using the same filter for both the image data and the modification function, followed by the division step, establishes an advantageous approach to image processing. This method acknowledges and addresses the changes introduced during the initial modification, ensuring they are not carried forward unchecked into the final image. This balanced approach to data alteration and correction maintains the integrity of the original image data. It ensures that the final image is not just a product of arbitrary modifications but a carefully adjusted representation that stays true to the original data. This is particularly important in OCT, where image accuracy and clarity are important for effective analysis and diagnosis. Optionally, the correction by dividing the result with the same filter applied to a modification (e.g. multiplication) function that was used to modify the initial complex-valued data, is subtracted from the initial data, to remove the artifacts in the filter. It can for example be used to remove imaging artifacts from optical coherence tomography data. In some examples, it is used to remove side- lobes of the point-spread function (apodization). Optionally, a point-wise, magnitude-dependent modification of image data is performed, each data point in the image data undergoing an alteration wherein the specific alteration applied to each data point is a function of its respective magnitude. This can ensure that the amplitude alteration is tailored to the unique intensity characteristics of each pixel within the image. The modification function can be used for weighting pixel contributions differently based on their intensity. Each data point can be altered based on its specific magnitude, allowing for precise and localized enhancement of the image. This approach provides a high degree of control over the image modification process. By modifying each data point based on its specific magnitude, the method enables localized adjustments to the image. This point-wise alteration means that the enhancement may not be applied uniformly across the entire image but is tailored to each pixel’s characteristics. Such a granular level of control is particularly beneficial in OCT, where precise image details are important. The specific alteration applied to each pixel is governed by the pixel’s magnitude. This can be important because the magnitude of a pixel in OCT data can convey important information about the underlying tissue structure or the presence of artifacts. By using the magnitude as a guiding parameter for modification, the method ensures that changes to the image data are relevant and meaningful, enhancing areas that require it while preserving areas that do not. Magnitude-dependent modifications allow for more effective reduction of noise and artifacts. This is because artifacts typically manifest as alterations in pixel magnitudes, and a tailored approach can specifically target and mitigate these alterations. Furthermore, since the modification is point-wise and dependent on the magnitude, it can preserve important details in the image. This is particularly important in medical imaging, where the loss of detail can lead to misinterpretation. Since side-lobes typically exhibit lower intensity compared to their corresponding main peaks, they may inherently receive a different “weight” during the modification process. This differential weighting is significant because it ensures that the side-lobes and main peaks are affected differently by the subsequent filtering process. In OCT imaging, for example, accurately distinguishing between main peaks and side-lobes is essential for clear and precise image interpretation. By applying variable weights based on the intensity magnitude of each pixel, the method ensures that side-lobes, which are lower in magnitude, are modified in a manner distinct from the higher magnitude main peaks. This tailored approach not only enhances the overall image quality by reducing noise and artifacts, but also preserves critical details in the image. Specifically, it helps in attenuating the side-lobes without disproportionately affecting the main peaks, thereby maintaining the integrity of the primary image information while effectively mitigating any distortions or artifacts typically associated with side-lobes. This careful handling of different image components provides for improved precision and adaptability. Optionally, the modification function is selected based on at least one of the following image processing / quality metrics: sparsity (L1 metric), Shannon entropy, or a power function. The choice of metric guides the modification process, ensuring that it is tailored to the specific needs and characteristics of the image data. The method may be adapted to different imaging requirements and characteristics, ensuring optimal enhancement for each specific scenario. For examples, sparsity (L1 Metric) can be used to enhance image clarity by emphasizing significant features while minimizing less important details. The L1 metric, or sparsity, measures the sum of the absolute values of the data points. By selecting a modification function based on this metric, the method prioritizes data points with higher magnitudes (more significant features) while reducing the impact of lower magnitude points (less significant details). This approach is particularly useful in images where the focus is on distinct features against a relatively uniform background, as it helps in isolating and enhancing these features while reducing noise. The L1 metric, focusing on sparsity, integrates into the process by initially summing either the square roots of the intensities (sum over all ^^0.5) or the absolute values of the amplitudes (sum over all |A|), where I represents intensityand A denotes amplitude or field strength, i.e., ^^ = |^^|2. This summation can formthe basis for deriving the weighting function. The actual weighting applied to each data point in the image can be then determined by the derivative of this function with respect to the intensity. In this context, since the derivative of ^^0.5is 0.5^^−0.5, the resulting weighting function becomes proportional to 1 / |A|. This method of weighting effectively emphasizes the phase information in the image while filtering out the amplitude variations. Such a derivative-based weighting ensures that the filtering process is finely attuned to the nuances of the image, enhancing image quality by preserving essential phase information while reducing amplitude- induced noise and artifacts. For example, Shannon entropy can be used to improve image data quality by accounting for the information content in the image. Shannon entropy measures the information content in the data, providing a sense of the complexity or randomness in the image. By using this metric to guide the selection of the modification function, the method can tailor its processing to the specific informational structure of the image. For instance, in images with a high degree of complexity or variability, the method can adapt to preserve this information while still enhancing overall clarity. While Shannon entropy can be associated with measuring information content, in the context of image processing such as OCT image processing, it can function differently. Here, Shannon entropy can be aligned more closely with the concept of sparsity, akin to the L1 metric, but with a unique formulation. This alternative application of Shannon entropy, rooted in the classical entropy formula, does not directly measure the informational content of the image. Instead, it aids in enforcing sparsity, which is a method of emphasizing significant image features while minimizing less prominent details. By utilizing Shannon entropy in this manner, the modification process in OCT image enhancement focuses on reducing complexity and randomness, leading to clearer and more defined images. Essentially, this method adapts to the specific structural characteristics of the image, not by assessing its informational content per se, but by effectively managing the distribution and prominence of key image features. For example, the power function can be used to enable precise control over the enhancement based on the magnitude of the image data. A power function modifies the data based on the magnitude raised to a certain power. This approach can amplify or attenuate features in the image based on their intensity. For example, in a scenario where subtle details are important, a power function can be chosen to amplify these minor variations, making them more prominent in the final image. Conversely, in situations where it is necessary to suppress overwhelming features, a different power function can be employed to achieve this effect. The application of a power function with the exponent I0.5(where I is the intensity of the image data) can be considered to be equivalent to employing the L1-metric in terms of amplitude. This can position the L1-metric as a special case within the broader spectrum of power function-based modifications. The power function, by allowing the exponent to vary, provides the capability to fine-tune the enhancement process according to the specific requirements of the image data. For instance, by raising the magnitude of the image data to the power of 0.5, the method aligns with the principles of the L1-metric, prioritizing sparsity and emphasizing significant image features while minimizing less important details. In some examples, the L1 metric is selected to minimize background. The background is then determined by filtering after applying the derivative of the metric as weighting function, that is dividing by the amplitude. The background is subtracted, the modified image now minimizing the L1 metric. In some examples, the modification function may be the derivative of the image processing metric function with respect to the intensity (i.e., magnitude squared). Optionally, the division process is performed with a modification of the division operation, wherein a regularization technique is employed involving computing a modified divisor, where the filtered reference is adjusted to include a stabilizing factor for increasing stability and robustness of the division process. A modified divisor can be computed where the filtered reference is adjusted to include a stabilizing factor. This factor increases the stability and robustness of the division process, ensuring more accurate and reliable compensation for the initial modifications. Division operations can be susceptible to instability, especially when the divisor approaches zero. A regularization technique can be employed to modify the division operation, thereby ensuring that it remains stable and robust under various conditions. Regularization can be performed by introducing additional information and / or constraints into the division process to prevent extreme values or erratic behavior. In some examples, the modification involves computing a divisor that is adjusted with a stabilizing factor. This factor is designed to prevent the divisor from becoming too small (approaching zero), which can lead to instability in the division process. For instance, a small positive constant may be added to the divisor. This adjustment is particularly important when dealing with real-world data that may contain low-intensity or near-zero values, as is often the case in OCT imaging. The stabilizing factor itself can be chosen based on the characteristics of the image data and the specific requirements of the filtering process. It should be sufficiently large to prevent instability but small enough to ensure that the compensation remains accurate and does not introduce significant distortions. By incorporating the stabilizing factor into the divisor, the division process becomes more stable. This means that it is less likely to produce extreme or undefined values, which can occur when dividing by values close to zero. The robustness of the process is also enhanced, meaning that the method can reliably handle a wider range of image data, including cases where certain regions of the image have very low intensity or are prone to noise. This robustness can be important for maintaining the integrity and quality of the processed image, ensuring that the compensation accurately reflects the original intent of the image modification without introducing new artifacts. Furthermore, the regularization of the division process can be important for maintaining the integrity of the image. This means that the processed image remains true to the original data in terms of structural and qualitative aspects, without suffering from distortions that can be introduced by unstable division operations. In the specific context of OCT, where images are often complex and contain a wide range of intensities, this stability can be particularly important. It ensures that the processed images are both accurate and reliable, providing a true representation of the scanned tissue or material. Optionally, the method further includes adding an offset to the division function used during the compensation process to prevent division by zero, wherein the offset comprises a value added to the division function at least in instances where the filtered reference value is zero. The compensation process can be further refined by adding an offset to the division function used during compensation. This is specifically to prevent division by zero scenarios. The offset comprises a value added to the division function in instances where the filtered reference value is zero, thereby maintaining the stability and accuracy of the compensation process. By adding an offset to the division function, the method ensures that the divisor never becomes zero. This small, added value (the offset) guarantees that the division operation remains well-defined and computationally stable. Computational errors from division by zero can lead to artifacts or distortions in the processed image. The offset ensures that these errors are avoided, thus maintaining the integrity and quality of the resulting image. The offset advantageously acts as a stabilizing factor in the division process. Even when the filtered reference value is very small or zero, the offset ensures that the division operation yields a finite, stable value. This approach is particularly advantageous in scenarios where the divisor may fluctuate around zero due to the characteristics of the image data or the filtering process. The value of the offset can be chosen to be small enough to not significantly affect the accuracy of the compensation but large enough to prevent computational instability. This balance can be important for ensuring that the offset fulfills its role without introducing its own distortions into the processed image. Optionally, the compensation process includes enhancing the filtered image data by employing a correction factor derived from the correction function. This enhancement may involve a division process that incorporates both the correction function and a stabilization element. The stabilization element may be introduced to ensure stable and robust division, particularly in scenarios where the correction function’s influence on the division process is minimal or approaches a critical threshold. This method can ensure a balanced and effective compensation for alterations introduced by the initial modification of the image data. Optionally, the compensation process includes multiplying the filtered image data with a correction term, c(x), divided by the sum of the square of the absolute value of the correction term, |c(x)|^2, and an offset, ε, represented as c*(x) / (|c(x)|^2 + ε), where c(x) is the correction function derived from the modification function and the filtering process, * indicates complex conjugation, and ε is a predetermined offset value introduced to prevent instability in the division operation. This value is typically based on the noise level of the image data. Optionally, the image data is complex-valued reconstructed OCT data. OCT data inherently contains complex values, where each data point holds both amplitude (magnitude) and phase information. The modification function can be configured to change the magnitude while preserving the phase. This can be important because the phase information in OCT data often contains valuable insights about the sample being imaged, such as depth information or material properties. OCT data has unique features, such as high resolution and depth- specific information, which are important in applications like medical imaging. The method’s tailored approach means it is fine-tuned to enhance these specific characteristics. For example, the filtering and modification techniques can be designed to optimize the depth resolution and clarity of OCT images, which can be important for accurate analysis in medical imaging. In some examples, initial complex-valued OCT image data is received. Each data point in this image is complex-valued, comprising a real and imaginary part. The image data undergoes a point-wise, magnitude-dependent modification, where each data point is altered based on its respective magnitude, while the phase of the data is maintained. This modification is executed by applying a function to the image data, which is selected based on image processing metrics such as sparsity (L1 metric), Shannon entropy, or a power function. In some examples, subsequently, a filtering process is applied to the modified image data. This process can be implemented either in the frequency domain, involving operations like the removal or attenuation of certain frequencies or the application of a window function, or as a convolution filter in the spatial domain of the image. The filtering process is adept at isolating background elements, signal side-lobes, and artifacts, which are then subtracted from the initial image data. The filtering process described plays an important role in enhancing image data, for example in optical coherence tomography (OCT). A two-step approach may be employed: first, employing the filtering process to isolate specific elements such as background noise, signal side-lobes, and artifacts from the initial image data; and then second, subtracting these isolated elements to cleanse the image. However, it will be appreciated that this is not the only approach possible with the method of this disclosure. The method also allows for the development of other filters that can more directly eliminate artifacts from the image data. Unlike standard linear filters, which typically adjust image data uniformly, these other filters can be specifically designed to target and remove artifacts based on their unique characteristics. This means that the filtering process is not just about altering the overall image data but may involve a more tailored / targeted approach. By designing filters that can identify and remove specific types of artifacts or noise, the method offers a more direct and efficient way of enhancing image quality. This adaptability in filter design, moving beyond standard linear filtering to more complex, artifact-specific filtering, significantly enhances the method’s utility in producing clear, high-quality images in various OCT imaging scenarios. In some examples, advantageously, compensation / correction is performed. The compensation / correction for alterations resulting from the initial modification of the image data can be achieved by determining a correction derived from applying a process analogous to the filtering process to the modification function. The filtered image data is then adjusted using this correction, effectively compensating for any artifacts introduced during the initial modification phase. The entire process can be designed to be iterative, where each iteration refines the image data progressively, continuing until a predetermined criterion is met or a fixed number of iterations has been executed. Optionally, the filtering process includes operations in the frequency domain, wherein the operations performed in the frequency domain include at least one of the following: removal and / or attenuation of certain frequencies to attenuate undesired signal components or application of a window function to modify the spectral characteristics of the signal. This approach allows for precise and targeted filtering, conducive to improved image quality. In the frequency domain, each frequency component corresponds to a specific feature or aspect of the spatial image. By manipulating these components, the method can effectively address specific types of noise, artifacts, or unwanted features that manifest as distinct frequencies. Certain types of noise or artifacts in OCT images may be associated with specific frequency ranges. By operating in the frequency domain, the method can selectively remove these frequencies, thereby eliminating the corresponding noise or artifacts from the image. In some cases, completely removing a frequency may not be desirable, as it may lead to loss of important image details. The method allows for the attenuation of frequencies instead, reducing their impact without completely erasing them. This nuanced approach can be important for maintaining the integrity of the image while enhancing its quality. Beyond noise removal, the method can also modify the frequency components to enhance certain image features. For example, boosting specific frequencies can sharpen image details or improve contrast, leading to clearer and more informative OCT images. Since OCT images can vary widely in terms of their frequency characteristics depending on the sample and imaging conditions, the ability to operate in the frequency domain provides a high level of adaptability. The method can be tailored to the specific requirements of different imaging scenarios. The precise control over frequency components makes it possible to optimize images for a variety of applications, whether it’s enhancing the visibility of fine structures in a biological sample or reducing noise in low-contrast images. This adaptability enhances the overall applicability of the method, making it a versatile tool in the field of OCT imaging. In some examples, a window function is applied. Various types of frequency domain operations may be performed. For example, band-pass filtering may be employed. Additionally or alternatively, adaptive filtering may be performed, where the frequency domain operations change based on the image content. Optionally, the filtering process is implemented as a convolution filter in a spatial domain of the image. Additionally or alternatively to the frequency domain approach, the method may also allow for the filtering process to be implemented as a convolution filter in the spatial domain of the image. This flexibility in the choice of filtering domain enables the method to be adapted to a range of image characteristics and processing requirements. Unlike frequency domain processing which involves transformations like Fourier Transform, spatial domain processing directly manipulates the pixel values in the image. This direct manipulation allows for more intuitive and immediate adjustments to the image’s visual characteristics and for small convolution kernels the filter may be performed faster. Convolution filters operate by applying a filter matrix (kernel) to each pixel and its neighbors. This method is highly effective for local enhancements, such as sharpening, blurring, edge detection, and noise reduction. Since the convolution process considers the spatial relationship between pixels, it is particularly adept at preserving or emphasizing these relationships, which can be important for maintaining the integrity of structural features in the image. Optionally, the filtering process is adapted to isolate background elements, signal side-lobes, and / or artifacts from the initial image data, and wherein the isolated background elements, signal side-lobes, and / or artifacts are subsequently subtracted from the initial image data to remove background noise, signal side-lobes, and / or artifacts. The method may be adapted to isolate specific elements such as background noise, signal side-lobes, and artifacts from the initial OCT image data. These isolated elements are then subtracted from the initial image data, effectively removing background noise, signal side-lobes, and artifacts, and thereby enhancing the overall image quality. By focusing on isolating specific elements like background noise, signal side-lobes, and artifacts, the method can target and eliminate these disturbances without affecting the vital details of the image. This selective removal can be important because indiscriminate filtering can lead to the loss of important image features, which are often necessary for accurate diagnosis and analysis in medical imaging applications like OCT. The filtering process may be fine-tuned to identify and separate the elements that are considered noise or artifacts. This can involve analyzing the image data to distinguish between the main signal and the background or side- lobes based on their distinct characteristics. For instance, background noise may have a different frequency profile compared to the actual image data. Once the unwanted elements are isolated, the method may then involve subtracting them from the original image. This subtraction can be carefully executed to ensure that only the identified noise and artifacts are removed, in order to maintain the integrity of the important image data. This is particularly important in OCT, where even minor details can be important for accurate interpretation. The result of this selective filtering and subtraction is a cleaner and clearer image. By effectively removing the elements that typically obscure the view or create confusion, the method enhances the overall clarity of the image. This clarity can be important in OCT imaging, where precise details can provide important information for diagnosis or analysis. Alternate examples can include different techniques for isolating these elements, such as machine learning-based methods. Optionally, the background noise is a fixed-pattern noise in the image data. Optionally, the method includes: defining a noise-identification function, which is configured to differentiate between essential image data and background noise within the image based on predetermined criteria; optimizing the noise-identification function through an iterative process that adjusts the function’s parameters to maximize its accuracy in distinguishing image data from noise; applying the optimized noise-identification function to the OCT image data to quantify the amount and location of background noise in the image; and subtracting the quantified background noise from the original OCT image data. A noise-identification function can be defined and optimized to differentiate between important image data and background noise. The function’s parameters can be adjusted iteratively to maximize accuracy in distinguishing between image data and noise. Once optimized, this function is applied to the OCT image data to quantify background noise, which is then subtracted from the original image data, enhancing image clarity. By defining and iteratively optimizing a noise-identification function, the method maximizes its accuracy in distinguishing between image data and noise. This precise quantification of background noise enables more effective subtraction and results in a cleaner image. The optimization of the noise-identification function can be an iterative process. It may involve adjusting the parameters of the function to maximize its accuracy in distinguishing between actual image data and noise. This dynamic optimization process allows for continuous improvement in the function’s performance, adapting to the nuances of the image data. As the function is refined through iterations, it becomes increasingly adept at identifying and quantifying noise, ensuring that the subtraction is as precise and effective as possible. Different OCT imaging scenarios may present various types of noise and artifacts. The method of the disclosure may be not limited to a specific type of noise or imaging condition; it can be adapted and optimized for a wide range of scenarios. This versatility makes the method broadly applicable across different OCT systems and various types of medical imaging needs, increasing its utility and effectiveness. Optionally, a single component along a given axis is determined for background identification, wherein the filter is applied by averaging each column or each row of pixels in the image data. The approach of a single component determination for background identification effectively simplifies the process of background noise identification, making it more computationally efficient while still effective. The filter can be applied by averaging each column (A-line or A-scan) or each row of pixels in the image data. The approach streamlines the identification and subtraction of background noise by determining a single component along a given axis for background identification, averaging each row or column of pixels. By concentrating on a single axis for background noise identification, the method can significantly reduce the computational complexity. Traditional methods may require analyzing the entire two-dimensional space of the image, involving complex calculations across both axes. In contrast, this approach reduces the problem to a one-dimensional analysis, greatly simplifying the mathematical operations involved. This reduction in complexity leads to faster processing times, making the method highly efficient, especially valuable in situations where rapid image processing can be important. This method may be particularly advantageous in scenarios where background noise exhibits consistency along one axis. In OCT imaging, certain types of artifacts or noise may predominantly appear in either the lateral (x) or depth (z) direction. In particular, if background noise / fixed-pattern noise is constant for all A-lines and phase stable for all A-lines it makes sense to use this approach, allowing for more robust and computationally efficient determining of the background. By targeting the axis along which the noise is uniform, the method can effectively reduce or eliminate these consistent noise components, thereby enhancing the overall image quality. In summary, this method may compromise of a weighted average of all pixels in the lateral direction of the B-scan (weighted with the modification function) to determine fixed pattern noise that is constant and phase stable for all A-lines. A weighted average typically includes a correction step by dividing with the sum of all weights. It mitigates artifacts that are often introduced by background subtraction in standard OCT signal processing and still effectively removes background. Despite its simplicity, this approach does not significantly compromise the quality of the final image. By averaging across a single axis, it effectively smooths out the noise while retaining the important features and details of the image. This balance is particularly important in medical imaging applications like OCT, where preserving image fidelity can be important for accurate diagnosis and analysis. Optionally, the Fourier transform is only partially computed. Optionally, the process is performed iteratively, with each iteration taking as its input the output of the previous iteration, thereby progressively enhancing the image data, wherein this iterative process is performed until a predetermined criterion is met or a fixed number of iterations has been executed. This progressive approach allows for continuous enhancement of the image data, leading to incrementally improved image quality. A repetitive cycle may be employed wherein each iteration applies the claimed processing techniques, including modification, filtering, and compensation, to the image data. This iterative process results in the progressive enhancement of the image data with each successive cycle. Each iteration can build upon the improvements made in the previous cycle, allowing for cumulative enhancements. This is particularly beneficial in dealing with complex or subtle image artifacts that may not be fully corrected in a single processing pass. Another significant advantage is the capability to tailor the enhancements to meet specific quality criteria. By adjusting the parameters and extent of processing in each iteration, the method can be fine-tuned to achieve the desired level of clarity and detail. This flexibility can be important in OCT imaging, where different diagnostic scenarios may require different image qualities. The iterative approach allows the processing method to adaptively respond to the changes in the image data. As the image is progressively enhanced, the method can adjust its processing parameters in response to the evolving characteristics of the image, ensuring that each iteration is optimized based on the current state of the image data. Iterative processing can converge towards an optimal level of image quality. As iterations proceed, the improvements in image quality can be monitored, and the process can be halted once the image meets the predetermined quality criteria or a fixed number of iterations is executed. This ensures that the image is not over-processed while still achieving the highest possible quality. According to an aspect, the invention provides for an optical coherence tomography system comprising a processor configured to perform the steps of: receiving an initial image data captured by employing an optical coherence tomography based imaging technique; modifying the image data using a modification function that is configured to change magnitude while maintaining phase of the image data, in order to obtain a modified image data, wherein the modification function is selected based on one or more image processing metrics; applying a filtering process to the modified image data using a filter to obtain a filtered image data; and performing a compensation on the filtered image data in order to compensate for alterations resulting from the initial modification of the image data prior to filtering, wherein the compensation is performed based on the modification function and the filter. Correcting errors / distortions introduced earlier involves a process of compensating for any distortions or alterations that occurred during initial stages of modification. This can be achieved by applying a reverse or inverse operation of the initial modification. For instance, if the initial step involved scaling the magnitude of the data, the correction step might involve scaling it back by a reciprocal factor. This approach ensures that while the desired enhancements or noise reductions are achieved, the fundamental characteristics of the original data, particularly the magnitudes, are preserved or restored to their original state. Advantageously, the accuracy and integrity of the processed data can be effectively maintained. Applying correction by dividing the result with the same filter used initially is an effective way to balance the alterations made during the filtering process. This corrective / compensation step may ensure that the final image retains its integrity, balancing enhancements with the preservation of the original data characteristics. Advantageously, the system is capable of maintaining image resolution while suppressing side lobes, achieving improved image quality without requiring hardware changes. According to an aspect, the invention provides for a non-transitory computer-readable medium storing instructions that, when executed by a processor, cause the processor to perform the method of the present disclosure. The point-wise multiplication function can significantly improve the process of background subtraction and PSF improvement. By adjusting the magnitude of each data point, the modification function can help in refining the overall quality of the image. It can be particularly effective in scenarios where specific characteristics of the image data need to be emphasized or suppressed. The modification function can be tailored, through one or more parameters, for different types of data and imaging requirements. This flexibility allows for optimization in various OCT imaging scenarios. The use of a non-linear function can offer more nuanced control over image processing, especially in dealing with complex data sets with varying intensity and contrast levels. The compensation step compensates / corrects the effect of alterations introduced by the initial modification of the image data. After the image data is modified and filtered, this step involves recalculating and adjusting the filtered image data based on the specific characteristics of both the modification function and the filter used. This correction ensures that any distortions or artifacts that were introduced during the modification phase are effectively neutralized, thereby restoring the integrity of the image data and enhancing its overall quality. This enables to achieve accurate and reliable results in various imaging applications, such as OCT or other imaging techniques such as radar, ultrasound, digital holography, etc. Artifacts in the image data, such as side-lobes in the point-spread function, can be effectively reduced without sacrificing the resolution of the image. It will be appreciated that, while the present invention has been primarily described in the context of exemplary Optical Coherence Tomography (OCT), the principles and methodologies detailed herein are not limited to this specific application. The inventive concepts embodied in this patent application, particularly those related to signal and / or image processing, are applicable to a wide range of fields that require computer-implemented data processing techniques. For instance, the techniques of modifying image data based on certain metrics, applying tailored filtering processes, and performing compensation for alterations introduced by initial modifications are equally pertinent to areas such as (synthetic aperture) radar signal processing, ultrasound imaging, and digital holography, and magnetic resonance imaging (MRI). In particular this applies to any coherent imaging technique. In these and various other fields, similar challenges are encountered, such as the need to isolate signal from noise, enhance the resolution of data, and effectively correct for distortions or artifacts introduced by earlier processing steps. Therefore, it is to be understood that the scope of the present invention encompasses any such modifications, variations, or equivalents that fall within the scope of the underlying principles disclosed and claimed herein. According to an aspect, the invention provides for a computer- implemented method and system for processing image data, comprising: receiving an initial image data captured by employing an imaging technique; modifying the image data using a modification function that is configured to change magnitude while maintaining phase of the image data, in order to obtain a modified image data, wherein the modification function is selected based on one or more image processing metrics; applying a filtering process to the modified image data using a filter so as to obtain a filtered image data; and performing a compensation on the filtered image data in order to compensate for alterations resulting from the initial modification of the image data prior to filtering, wherein the compensation is performed based on the modification function and the filter. According to an aspect, the invention provides for a computer- implemented method and system for processing complex-valued image data, comprising: modifying the image data based on selected processing metrics to change magnitude while preserving phase; applying a filtering process to the modified image data; and performing a compensation on the filtered image data to correct alterations caused by the initial modification, based on the modification function and filter characteristics. The methods and systems can be applicable across various imaging techniques including Optical Coherence Tomography (OCT), radar, ultrasound, and digital holography. It will be appreciated that in OCT, data acquisition typically occurs in the frequency space and may require resampling or other techniques for processing. The intermediate result of this process is an image comprising complex values, achieved after the reconstruction phase. It will be appreciated that any of the aspects, features and options described in view of the computer implemented method apply equally to the system and the described device. It will also be clear that any one or more of the above aspects, features and options can be combined. BRIEF DESCRIPTION OF THE DRAWING The invention will further be elucidated on the basis of exemplary embodiments which are represented in a drawing. The exemplary embodiments are given by way of non-limitative illustration. It is noted that the figures are only schematic representations of embodiments of the invention that are given by way of non-limiting example. In the drawing: Fig. 1 shows a schematic diagram of an embodiment of a method; Fig. 2 shows exemplary results of processed OCT image data; and Fig. 3a-c show exemplary image processing results. DETAILED DESCRIPTION Fig. 1 shows a schematic diagram of an embodiment of a method 100 for processing image data in optical coherence tomography. In a first step 101, an initial image data captured by employing an optical coherence tomography based imaging technique is received. In a second step 102, the image data is modified using a modification function that is configured to change magnitude while maintaining phase of the image data, in order to obtain a modified image data, wherein the modification function is selected based on one or more image processing metrics. In a third step 103, a filtering process is applied to the modified image data using a filter so as to obtain a filtered image data. In a fourth step 104, a compensation is performed on the filtered image data in order to compensate for alterations resulting from the initial modification of the image data prior to filtering, wherein the compensation is performed based on the modification function and the filter. The method can provide for a robust artifact removal and / or point- spread function shaping in e.g. scanning and parallelized Fourier-domain optical coherence tomography. In some examples, the method involves a process of modifying the magnitude of each data point in complex-valued data, wherein the modification is applied in a manner that specifically depends on the magnitude of the data point itself, while importantly preserving the phase information of the data. This can ensure that the essential characteristics of the original data are not distorted while enhancing the data’s clarity. Various filtering techniques may be employed. In some examples, linear filtering with Fourier transform techniques is employed. Following the initial modification, the data undergoes a linear filtering process. This process may be implemented by computing either the inverse or the direct discrete / fast Fourier transform across any number of dimensions within the data set. The filtering process may not only modify the magnitudes but may also implement the necessary inverse / discrete Fourier transform operations, thereby refining the data quality further. The correction of any errors introduced during the initial modification and filtering stages provides for significant benefits. Advantageously, this correction process is designed to revert the data to its original magnitude, ensuring that the integrity of the initial data is preserved while still benefiting from the enhancements of the preceding steps. In some examples, the non-linear filtering method is employed in the context of optical coherence tomography. In this application, the method is particularly effective in removing imaging artifacts, thereby significantly improving the quality of OCT images. This enhancement can be important for accurate analysis in medical imaging. In some examples, the modification of the magnitude is executed by multiplying the data with a function that is explicitly dependent on the magnitudeof the complex-valued data. An example of such a function is Γ′(I(x)) = I(x)γ whereI is the magnitude squared of the complex-valued data, i.e., I(x) = |U|2(x), with U(x) the complex valued data and x the pixel index in the image. Γ′(I(x))is the derivative of Γ(I(x)) with respect to I. This precise manipulation offers a tailored approach to data enhancement, allowing to adapt to specific requirements of the imaging process. In some examples, a metric is selected, and the derivative is then used for weighting. For instance ^^^^may be selected, and the weighting can be based on In some examples, only a single component along a given axis is determined, and where the Fourier transform is only partially computed. These approaches offer more focused and efficient data processing techniques, which can be particularly advantageous in scenarios where specific data characteristics need to be emphasized or when computational resources are limited. In some examples, a method of identifying and removing unwanted noise is employed. A function may be used which scrutinizes each segment of the image to determine whether it constitutes significant content or merely background noise. This function acts as a critical tool in distinguishing between essential image details and extraneous noise elements. Then, the noise identification process can be optimized. This entails an optimization technique, which systematically adjusts the parameters of the noise-identifying function. The objective of this optimization is to enhance the function’s precision in differentiating vital (i.e. not noise, artefacts, etc.) image components from the noisy background. Essentially, this process fine-tunes the filter, ensuring its efficacy in accurately isolating noise elements within the image. Upon achieving an optimized noise identification function, this function can be applied across the entire image. This application is comprehensive, assessing each part of the image to ascertain the extent of noise present. It is a meticulous process that quantifies the noise in every segment of the image, preparing the data for the subsequent noise removal phase. With a clear understanding of the noise distribution within the image, the method can then focus on eliminating this identified noise. This step is executed by subtracting the quantified noise elements from the image, effectively erasing the unwanted noise. This subtraction results in a significant enhancement of the image’s clarity. The resultant image is not only cleaner but also more precise, making it particularly valuable in applications where image clarity is important. In some examples, an adjustable filter can be used that is adept at both identifying and removing noise from images. The noise-identification function that is configured identify the background noise in the image data may be optimized by finding the function that minimizes or maximizes a specific metric. This is akin to adjusting the parameters of the noise-identification function to make it as accurate as possible in distinguishing between noise and essential image data. Then, the optimized function can be applied to quantify noise. More particularly, the noise- identification function may be applied to the original image data to quantify the amount and location of background noise. The image data may for instance be a constructed OCT image data. Then, the quantified background noise is subtracted from the original image, resulting in an enhanced image with reduced background interference. In some examples, the process of background identification can be significantly simplified and made computationally efficient by focusing on a single component along a given axis of the image data. This approach can be particularly advantageous in Optical Coherence Tomography (OCT) where the background often exhibits consistent characteristics across specific dimensions of the image. In OCT images, the background often manifests uniformly across each column (or row, depending on the axis of interest). This uniformity allows for a simplified approach to background subtraction, as it reduces the need for complex computations across the entire image. The method may involve identifying a single representative component for each column (or row). This can be as simple as determining a single pixel value for each column that best represents the background. This approach eliminates the need for performing a full Fourier transform, which is computationally intensive and time-consuming. Instead of a Fourier transform, the method may use a simpler operation, such as averaging, to determine the background component. By averaging all pixels in a column, the method effectively captures the predominant background signal, which is assumed to be consistent across the column. To further refine this process, the method may incorporate a weighted averaging approach. Here, pixels are weighted differently based on their brightness – darker pixels might be given more weight and brighter pixels less. This weighting scheme addresses the issue of high reflections or variations within the sample that might otherwise introduce new artifacts into the image. Median-based approaches for background subtraction may be computationally intensive as they require sorting and processing the entire data set. The weighted averaging method offers a similar level of effectiveness but is more computationally efficient, making it a more practical solution for real-time or high-volume image processing. The single component analysis enables easy implementation and integration into existing imaging systems such as OCT systems. It offers a quick and efficient way of subtracting background noise, enhancing the overall quality of the OCT images. This approach is particularly beneficial in scenarios where computational resources are limited or rapid processing is required. By focusing on a single component along a given axis and employing for example weighted averaging, it achieves significant computational efficiency while maintaining or even enhancing the accuracy and quality of the background subtraction process. This method is particularly advantageous over traditional Fourier transform or median-based approaches, offering an improved OCT image processing. In some cases, an assumption can be made that background noise in OCT images primarily depends on the depth direction (z-axis) and not on the lateral scanning direction (x-axis). This assumption is based on common characteristics observed in many OCT systems and imaging scenarios, where the background noise exhibits a significant dependency on the depth of the image rather than its lateral dimensions, particularly for point scanning OCT systems. This initial assumption may be gradually relaxed to allow for a more generalized process. For instance, certain frequency components may be considered to depend on the lateral direction (x-axis), but not to the same extent as they would if the noise depended equally on both x and z dimensions. It may be assumed that each function at a specific location in the image is only influenced by other functions at that same location. This simplification involves considering partial derivatives of the image data with respect to both lateral and depth directions, but ultimately leads to a conclusion where the intensity at a point depends only on the data at that point, not on other locations. This approach can streamline the derivation process and effectively simplify the computational requirements without compromising the accuracy of the result. A balance can be obtained between the specificity of the initial assumption and the flexibility required for broader application scenarios. In some examples, instead of considering a sum over all lateral and depth dimensions, a solution can be found that is consistent across the lateral dimension but varies with depth. This approach leads to obtaining a function that remains dependent on the depth dimension, fitting the initial premise of background noise dependency. Fig. 2a-c show exemplary results of processed OCT image data. The images are obtained with a swept-source based, scanning polarization-sensitive OCT system, acquired with an endoscope. In this example, the employed computer implemented method is based on a model of background and artifacts in Fourier-Domain Optical Coherence Tomography (FD- OCT), that relates the artifact-free spectral data S0 (k, x) and the measured spectral data S(x, k) by S(x, k) = R (k, x) S0 (k, x) + o (k, x) . Here, R (k, x) is a reference determining thespectral shape, o (k, x) is an additive offset, k describes the wavenumber and x is the (cumulative) scanning or lateral position. Without further assumptions, the three signals R, S0, and o may not be separated. Based on the assumption that R(k, x) and o (k, x) vary slowly with x compared to S0, several approaches to reduce the background are possible. In an advantageous example, a method is employed that is based on non-linear filtering to separate the background and the interference spectrum. This is compared to background subtraction of the lateral average spectrum / complex A-scan, and linear filter (lateral filtering). Selected results The method was first evaluated with data from endoscopic images of a polarization-sensitive (PS) swept-source OCT. Fig.2 shows the average intensity B-scans (left column), zoomed regions, and the phase difference of one polarization channel to the corresponding data without any background subtraction (right column). Subtracting the average spectrum (first row) does not reduce all artifacts, and even creates new horizontal lines from strongly reflecting structures. A lateral filter (second row) can effectively remove the artifacts, but results in broadening / side bands of the lateral PSF. The proposed non- linear filter with axial PSF shaping (bottom row) prevents the degradation of the lateral PSF and creates a clean, high-quality image data. At the same time the phase is maintained in all scattering structures, as apparent from the low phase difference. In regions of artifact, the phase is changed, as is expected, since the true phase is essentially unknown. Finally, the phase in the noise may change as a result from background subtraction. Fig.3a-c show exemplary image processing results. The effect on full-field FD- OCT data, acquired at 100 MHz A-scan rate, has been investigated. Full-field FD-OCT data with background signal from multimode illumination is shown in fig.3a. Simulated with additional relative intensity noise is shown in fig.3b. The proposed background subtraction (fig.3c) applied to fig.2b establishes good imaging quality despite the prominent noise sources. By sending the light through a multi-mode fiber, spatial coherence was destroyed thereby reducing the effect of multiple scattered light and flattening the illumination field but it also creates artifacts with low spatial frequencies. The obtained image is seen in Fig.3a. Severe relative intensity noise (RIN) has been added by multiplying raw data for each wavelength with a normal-distributed random number with an expectation value of 1 and a standard deviation of 0.1, which assumes a white spectrum for the RIN. This almost completely removed any remaining image information, see fig.3b. The proposed method completely removed the effect of the multimode fiber and the RIN. It uses the fact that all lateral pixels, i.e., all identical spectral components, are equally affected by the RIN. Thus, based on simulations, it has been determined that algorithms for line- field and full-field FD-OCT can be used for performing self-balancing, effectively eliminating relative intensity noise (RIN) and thus eliminating the need for balanced detection with a second detector or camera. The method provides for a robust approach to separate background / artifacts from actual OCT signals with the advantage of maintaining signal levels, reducing side-lobes, as well as shaping the axial point-spread function. At the same time, the method does not introduce other artifacts for most specimen. With the approach having only minimal impact on processing time, the method can be the ideal choice for many imaging scenarios. In the following, an exemplary derivation of a filter function is given. In a simplified scenario, an acquired (complex-valued) OCT imageU0(x, z) after reconstruction is considered, that is subject to background noise O(z). It is assumed that this background noise only depends on z and not on x. This assumption can be relaxed later on. To obtain a cleaned image, it is desired to subtract the noise and obtain the improved image U(x,z): ^^(^^, ^^) = ^^0(^^, ^^) − ^^(^^),Task is to determine O(z) such that it optimizes (minimizes or maximizes) a certain metric given by is performedwith respect to O(z), which gives ∂S ∂O z= ∑∂S∂I(x, z) ∂U(x, z)x,z( ) ( ), ( )∂I x, z ∂U x, z∂O(x) where it is already assumed that each function at a specific location isonly impacted by the other functions at the same location, e.g., I(x, z) only dependson U(x, z) with the same x and z.One can obtain the following intermediate results, ∂S dΓ ∂I(x, z)= | , dI I=I(x,z) in the following referred to as Γ′(I(x, t)). Further, which is to be understood as the Wirtinger derivative, and ∂U(x, z)∂O(x)= −1.Combining the above, it follows: To have the expression equal, one can take one solution that is equal in all points instead of the sum over all z, thus This is the weighted average of the original image, where it is assumedthat I ≈ I0 (with I0 = |U0|2when computing Γ′(I). The average over x can also be written as or even as a filter with w(k) = δ(k),i.e., it is 1 for the DC component (k=0) and 0 otherwise. In this caseO(x, z) will be independent of x, but this would no longer be the case if wegeneralize to arbitrary choices of w(k). It is possible to generalize from here and replace w(k)with an arbitrary function and generally x and z can represent any number of arbitrary variables in any number of dimensions: with w(k) being a Filter function and z replaced with a more generic variable t. Advantageously, imaging (e.g. OCT imaging) can have better imaging quality by better background subtraction, or a sharper / narrower PSF. Production costs and effort could be decrease, for example, a manual selection of suitable light sources would possible not be necessary. Also, a comparable image quality can be obtained at lower costs for the initial light source. Images can be improved and better separated from artifacts and fixed pattern noise. Particularly for some OCT systems, such as Full-field Fourier domain systems when imaging object with high signal-to-noise-ratio this becomes important. The method enables improved background subtraction. Additionally or alternatively, the resolution can be increased. The method may thus focus on the subtraction of background noise and / or the optimization of image clarity / resolution through filtering techniques. Background noise may be determined and subtracted from the reconstructed OCT image. The objective may be to optimize a specific metric, either by minimization or maximization. The process may involve differentiating with respect to the noise-identification function O(z) and using various intermediate results to construct the final expression for the noise-identification function O(z). Fourier transforms can be applied, both direct and inverse. The Wirtinger derivative can be utilized in the calculation process. A weighted average of the original image can be employed, with assumptions about the image intensity. This average is further refined through Fourier transforms, leading to a more precise and effective background noise subtraction. In a generalized form, this method extends to arbitrary dimensions and variables, using a filter function. The versatility of this approach allows for the application of various filter functions, thereby enabling customized solutions based on specific OCT imaging requirements. This adaptability is important for optimizing image quality across different OCT systems and imaging conditions. In some examples, the goal is to remove the background of an image. A metric may be selected that is to be minimized / maximized in the resulting background-subtracted image. This metric can be (based on) the sum of all Gamma(I(x, z)) for all x, z (i.e. pixels) in the image and this would be then the metric (a single value for the entire image). This metric can be chosen based on the image that is to be obtained. Based on this metric, a weighting function can be chosen, which can be the derivative Gamma' = dGamma / dI evaluated for each pixel. This can be applied, i.e., multiplied pixel-by-pixel. After filtering and correcting, the background that is removed from the initial image is obtained. The resulting image after the removal then can minimize (or theoretically maximize) the image quality metric. Optionally, this metric may also be used to determine which features are to be emphasized to get a better estimation of the background after filtering. It will be appreciated that the method may include computer implemented steps. All above mentioned steps can be computer implemented steps. Embodiments may comprise computer apparatus, wherein processes performed in computer apparatus. The invention also extends to computer programs, particularly computer programs on or in a carrier, adapted for putting the invention into practice. The program may be in the form of source or object code or in any other form suitable for use in the implementation of the processes according to the invention. The carrier may be any entity or device capable of carrying the program. For example, the carrier may comprise a storage medium, such as a ROM, for example a semiconductor ROM,hard disk, flash memory, or SSD. Further, the carrier may be a transmissible carrier such as an electrical or optical signal which may be conveyed via electrical or optical cable or by radio or other means, e.g. via the internet or cloud. Some embodiments may be implemented, for example, using a machine or tangible computer-readable medium or article which may store an instruction or a set of instructions that, if executed by a machine, may cause the machine to perform a method and / or operations in accordance with the embodiments. Various embodiments may be implemented using hardware elements, software elements, or a combination of both. Examples of hardware elements may include processors, microprocessors (e.g. CPU, GPU, TPU, etc.), circuits, application specific integrated circuits (ASIC), programmable logic devices (PLD), digital signal processors (DSP), field programmable gate array (FPGA), logic gates, registers, semiconductor device, microchips, chip sets, et cetera. Examples of software may include software components, programs, applications, computer programs, application programs, system programs, machine programs, operating system software, mobile apps, middleware, firmware, software modules, routines, subroutines, functions, computer implemented methods, procedures, software interfaces, application program interfaces (API), methods, instruction sets, computing code, computer code, et cetera. Herein, the invention is described with reference to specific examples of embodiments of the invention. It will, however, be evident that various modifications, variations, alternatives and changes may be made therein, without departing from the essence of the invention. For the purpose of clarity and a concise description, features are described herein as part of the same or separate embodiments, however, alternative embodiments having combinations of all or some of the features described in these separate embodiments are also envisaged and understood to fall within the framework of the invention as outlined by the claims. The specifications, figures and examples are, accordingly, to be regarded in an illustrative sense rather than in a restrictive sense. The invention is intended to embrace all alternatives, modifications and variations which fall within the scope of the appended claims. Further, many of the elements that are described are functional entities that may be implemented as discrete or distributed components or in conjunction with other components, in any suitable combination and location. In the claims, any reference signs placed between parentheses shall not be construed as limiting the claim. The word ‘comprising’ does not exclude the presence of other features or steps than those listed in a claim. Furthermore, the words ‘a’ and ‘an’ shall not be construed as limited to ‘only one’, but instead are used to mean ‘at least one’, and do not exclude a plurality. The term “and / or” includes any and all combinations of one or more of the associated listed items. The mere fact that certain measures are recited in mutually different claims does not indicate that a combination of these measures cannot be used to an advantage.

Claims

Claims 1. A computer-implemented method for processing image data in optical coherence tomography, comprising: receiving an initial image data captured by employing an optical coherence tomography based imaging technique; modifying the image data using a modification function that is configured to change magnitude while maintaining phase of the image data, in order to obtain a modified image data, wherein the modification function is selected based on one or more image processing metrics; applying a filtering process to the modified image data using a filter so as to obtain a filtered image data; and performing a compensation on the filtered image data in order to compensate for alterations resulting from the initial modification of the image data prior to filtering, wherein the compensation is performed based on the modification function and the filter.

2. The method of claim 1, wherein performing the compensation includes: determining a correction derived from applying a process analogous to the filtering process to the modification function; and adjusting the filtered image data using this correction to address alterations caused by the initial modification of the image data.

3. The method of claim 1 or 2, wherein performing the compensation includes: applying the same filter used in the filtering process to the modification function to obtain a filtered reference; and dividing the filtered image data by a term indicative of and / or based on the filtered reference in order to compensate for alterations introduced by the initial modification of the image data.

4. The method of any one of the preceding claims, wherein a point-wise, magnitude-dependent modification of image data is performed, each data point in the image data undergoing an alteration wherein the specific alteration applied to each data point is a function of its respective magnitude.

5. The method of any one of the preceding claims, wherein the modification function is selected based on at least one of the following image processing metrics: sparsity (L1 metric), Shannon entropy, or a power function.

6. The method of any one of the preceding claims 3-5, wherein the division process is performed with a modification of the division operation, wherein a regularization technique is employed involving computing a modified divisor, where the filtered reference is adjusted to include a stabilizing factor for increasing stability and robustness of the division process.

7. The method of any one of the preceding claims 3-6, further including adding an offset to the division function used during the compensation process to prevent division by zero, wherein the offset comprises a value added to the division function at least in instances where the filtered reference value is zero.

8. The method of any one of the preceding claims, wherein the image data is complex-valued reconstructed OCT data.

9. The method of any one of the preceding claims, wherein the filtering process includes operations in the frequency domain, wherein the operations performed in the frequency domain include at least one of the following: removal and / or attenuation of certain frequencies to attenuate undesired signal components or application of a window function to modify the spectral characteristics of the signal.

10. The method of any one of the preceding claims 1-8, wherein the filtering process is implemented as a convolution filter in a spatial domain of the image.

11. The method of any one of the preceding claims, wherein the filtering process is adapted to isolate background elements, signal side-lobes, and / or artifacts from the initial image data, and wherein the isolated background elements, signal side- lobes, and / or artifacts are subsequently subtracted from the initial image data to remove background noise, signal side-lobes, and / or artifacts.

12. The method of claim 11, including: defining a noise-identification function, which is configured to differentiate between essential image data and background noise within the image based on predetermined criteria; optimizing the noise-identification function through an iterative process that adjusts the function’s parameters to maximize its accuracy in distinguishing image data from noise; applying the optimized noise-identification function to the OCT image data to quantify the amount and location of background noise in the image; and subtracting the quantified background noise from the original OCT image data.

13. The method of claim 11 or 12, wherein a single component along a given axis is determined for background identification, wherein the filter is applied by averaging each column or each row of pixels in the image data.

14. The method of any one of the preceding claims, wherein the process is performed iteratively, with each iteration taking as its input the output of the previous iteration, thereby progressively enhancing the image data, wherein this iterative process is performed until a predetermined criterion is met or a fixed number of iterations has been executed.

15. An optical coherence tomography system comprising a processor configured to perform the steps of: receiving an initial image data captured by employing an optical coherence tomography based imaging technique; modifying the image data using a modification function that is configured to change magnitude while maintaining phase of the image data, in order to obtain a modified image data, wherein the modification function is selected based on one or more image processing metrics; applying a filtering process to the modified image data using a filter to obtain a filtered image data; and performing a compensation on the filtered image data in order to compensate for alterations resulting from the initial modification of the image dataprior to filtering, wherein the compensation is performed based on the modification function and the filter.

16. A non-transitory computer-readable medium storing instructions that, when executed by a processor, cause the processor to perform the method of any preceding claim 1-14.

Images