Jon Yngve Hardeberg - Academia.edu (original) (raw)

Papers by Jon Yngve Hardeberg

arXiv (Cornell University), May 24, 2019

This work covers multiple aspects of overt visual attention on 3D renders: measurement, projectio... more This work covers multiple aspects of overt visual attention on 3D renders: measurement, projection, visualization, and application to studying the influence of material appearance on looking behaviour. In the scope of this work, we ran an eye-tracking experiment in which the observers are presented with animations of rotating 3D objects. The objects were rendered to simulate different metallic appearance, particularly smooth (glossy), rough (matte), and coated gold. The eye-tracking results illustrate how material appearance itself influences the observer's attention, while all the other parameters remain unchanged. In order to make visualization of the attention maps more natural and also make the analysis more accurate, we develop a novel technique of projection of gaze fixations on the 3D surface of the figure itself, instead of the conventional 2D plane of the screen. The proposed methodology will be useful for further studies of attention and saliency in the computer graphics domain.

2019 IEEE/CVF International Conference on Computer Vision Workshop (ICCVW), 2019

Cracks on a painting is not a defect but an inimitable signature of an artwork which can be used ... more Cracks on a painting is not a defect but an inimitable signature of an artwork which can be used for origin examination, aging monitoring, damage identification, and even forgery detection. This work presents the development of a new methodology and corresponding toolbox for the extraction and characterization of information from an image of a craquelure pattern. The proposed approach processes craquelure network as a graph. The graph representation captures the network structure via mutual organization of junctions and fractures. Furthermore, it is invariant to any geometrical distortions. At the same time, our tool extracts the properties of each node and edge individually, which allows to characterize the pattern statistically. We illustrate benefits from the graph representation and statistical features individually using novel Graph Neural Network and hand-crafted descriptors correspondingly. However, we also show that the best performance is achieved when both techniques are merged into one framework. We perform experiments on the dataset for paintings origin classification and demonstrate that our approach outperforms existing techniques by a large margin.

ArXiv, 2019

Deep learning algorithms have demonstrated state-of-the-art performance in various tasks of image... more Deep learning algorithms have demonstrated state-of-the-art performance in various tasks of image restoration. This was made possible through the ability of CNNs to learn from large exemplar sets. However, the latter becomes an issue for hyperspectral image processing where datasets commonly consist of just a few images. In this work, we propose a new approach to denoising, inpainting, and super-resolution of hyperspectral image data using intrinsic properties of a CNN without any training. The performance of the given algorithm is shown to be comparable to the performance of trained networks, while its application is not restricted by the availability of training data. This work is an extension of original "deep prior" algorithm to HSI domain and 3D-convolutional networks.

In the last couple of decades, hyperspectral, multispectral, and multimodal (HMM) imaging has eme... more In the last couple of decades, hyperspectral, multispectral, and multimodal (HMM) imaging has emerged as an essential tool in various fields of science, medicine, and technology. Compared to integrated broad-band information as, e.g., present in RGB images, HMM imaging strives to acquire a multitude of specific narrow bands of the electromagnetic spectrum in order to solve specific detection or analysis tasks. HMM research is interested in studying light-matter interaction in a wide range of wavelengths from the high energy radiation down to Terahertz radiation (sub-millimeter waves). Furthermore, combining spectral data captured using different imaging modalities can unveil additional information of the scene that is not revealed solely by each of the individual imaging modalities. The workshop intended to connect researchers from different disciplines that involve HMM imaging and analysis. Even though there are very different approaches towards HMM imaging research and application...

Optics Express, 2021

A colour appearance model based on a uniform colour space is proposed. The proposed colour appear... more A colour appearance model based on a uniform colour space is proposed. The proposed colour appearance model, ZCAM, comprises of comparatively simple mathematical equations, and plausibly agrees with the psychophysical phenomenon of colour appearance perception. ZCAM consists of ten colour appearance attributes including brightness, lightness, colourfulness, chroma, hue angle, hue composition, saturation, vividness, blackness, and whiteness. Despite its relatively simpler mathematical structure, ZCAM performed at least similar to the CIE standard colour appearance model CIECAM02 and its revision, CAM16, in predicting a range of reliable experimental data.

2019 IEEE/CVF International Conference on Computer Vision Workshop (ICCVW), 2019

Deep learning algorithms have demonstrated state-ofthe-art performance in various tasks of image ... more Deep learning algorithms have demonstrated state-ofthe-art performance in various tasks of image restoration. This was made possible through the ability of CNNs to learn from large exemplar sets. However, the latter becomes an issue for hyperspectral image processing where datasets commonly consist of just a few images. In this work, we propose a new approach to denoising, inpainting, and superresolution of hyperspectral image data using intrinsic properties of a CNN without any training. The performance of the given algorithm is shown to be comparable to the performance of trained networks, while its application is not restricted by the availability of training data. This work is an extension of original "deep prior" algorithm to hyperspectral imaging domain and 3D-convolutional networks.

Color and Imaging Conference, 2018

Color Imaging: Device-Independent Color, Color Hardcopy, and Graphic Arts VI, 2000

Due to the increasing popularity and affordability of color imaging devices, color characterizati... more Due to the increasing popularity and affordability of color imaging devices, color characterization for these devices becomes an important subject. In other words, a set of color profile(s) needs to be generated for each device to transform the device dependent color space to a device independent one. This paper will concentrate on color characterization of scanners. Up to now, most scanner characterization has been optimized for photographic materials only. However, the color characterization inevitably depends on the media type and the printing mechanism that produced the target. In this work, a color scanner was characterized based on six different scan target types. For each target type, a color profile from RGB to CIELAB space was generated using polynomial regression method and a series of color conversions were conducted using color profiles optimized for itself as well as other target types. The accuracy of color transformation was evaluated. Results proved that to achieve the best color reproduction quality, it's necessary to characterize a color scanner based on specific scan target type. Considering the system level memory limitation, a compromise could be made without great quality degradation, using an average or unified profile for some target types. Nevertheless, a scanner should at least have two profiles: one for photo and the other for halftoning printing process.

SPIE Proceedings, 1999

To achieve high image quality throughout a digital imaging system, the first requirement is to en... more To achieve high image quality throughout a digital imaging system, the first requirement is to ensure the quality of the device that captures real-world physical images to digital images, for example a desktop scanner. Several factors have influence on this quality: optical resolution, bit depth, spectral sensitivities, and acquisition noise, to mention a few. In this study we focus on the colorimetric faculties of the scanner, that is, the scanner's ability to deliver quantitative device-independent digital information about the colors of the original document. We propose methods to convert from the scanner's device-dependent RGB color space to the standard device-independent color space sRGB. The methods have been evaluated using several different desktop scanners. Our results are very good: mean CIELAB (Delta) E*ab color errors as low as 1.4. We further discuss advantages and disadvantages of a digital color imaging system using the sRGB space for image exchange, compared to using other color architectures.

Lecture Notes in Computer Science, 2012

Based on previous research on super-resolution and colour correction by example, the present proj... more Based on previous research on super-resolution and colour correction by example, the present project employs such techniques towards a high-quality reclamation of lost art. The aim is to produce high-quality images of nowadays destroyed or missing paintings using the correspondence between similar artworks available in both low and high quality images. Several approaches for both super-resolution and colour correction techniques have been studied, implemented and tested to result to the most efficient and appropriate to be used. This project is an attempt that has never been done before and the successful reclamation of lost art initialises a new area of colour imaging applications in fine art. It reveals unlimited possibilities in the domain and establishes the potential of further attempts in this direction. iii

SPIE Proceedings, 1998

... Genevi eve Dardier, Jon Yngve Hardeberg and Hans Brettel ... Some manufacturers seem to take ... more ... Genevi eve Dardier, Jon Yngve Hardeberg and Hans Brettel ... Some manufacturers seem to take the necessary steps in order to signi cantly improve the quality of the pho-tographs by introducing a new type of image lters,1,2 using LCD technology, to correct areas on ...

A new method for the colorimetric characterization of a prin ter is proposed. It can also be appl... more A new method for the colorimetric characterization of a prin ter is proposed. It can also be applied to any other type of digital image reprodu ction device. The method is based on a computational geometry approach. It use s a 3D triangulation technique to build a tetrahedral partition of the pri nter color gamut volume and it generates a surrounding structure enclosing the defin ition domain. The characterization provides the inverse transformation from the device-independent color space CIELAB to the device-dependentcolor space CMY, takin g into account both colorimetric properties of the printer, and color gamut map ping.

Image reproduction suffers from several limitations in a color management system. In this paper, ... more Image reproduction suffers from several limitations in a color management system. In this paper, we have investigated artifacts resulting from the inherent characteristics of the color transformations by interpolation in three-dimensional look-up tables, and the unavoidable measurement noise of the color measurement done during device profiling. In our experiments, images were manipulated using three interpolation methods, and five levels of random noise. Psychophysical experiments were conducted to evaluate the quality of the reproduced images. Finally the experimental data were collected to analyze the color transformations, and test the performance of two color image difference metrics in this context. Introduction There has been increased demand to reproduce images using heterogeneous devices and media such as digital cameras, displays and printing systems. The employment of different color imaging devices results in a common problem that each device produces color differently. For example, to use the same values based on the device primaries, such as RGB for a display, would reproduce different colors by different printing systems. Hence, users are lacking color predictability and consistency to reproduce color images across different imaging media. This has been a driving force for the industry to develop technology to achieve successful cross-media color reproduction. Numerous attempts for the development of color management systems have been made to satisfy different image reproduction tasks from one medium to another. The most widely used systems are those based on the International Color Consortium (ICC) specifications. The ICC specification version 4 (1) provides definitions of color management architecture, profile format, and data structure. A typical ICC-based color management system consists of four basic components: profile, profile connection space (PCS), rendering intent, and color management module (CMM). A profile is a standard formatted file describing the device characterization, which defines the relationship between a device's control signals and the actual color that those signals produce. The ICC profile often employs multi-dimensional look- up tables (LUTs) to store the desired values. A process known as device characterization (or profiling) serves for this purpose, which provides a reliable way for color communications between media, and it is sufficient in simple applications with well-specified viewing conditions. The CMM is simply a color engine or processing engine, which is typically built into operating system, application or output device. The CMM performs all calculations needed to translate from the color space of one device to that of another. Although the ICC specifies the format of color profiles and to some extent the types of transforms that must be taken place to match colors between profiles, much of the process is left up to the imagination of the CMM creators. While it is difficult to specifically evaluate a vendor's commercial secrets, CMMs can certainly be evaluated based on the results they produce, both objectively by analysis of measurements and subjectively for pleasing contents. However, because different profiling applications will generate slightly different profiles from the same set of measurement data, the choice of CMM makes far less difference than the choice of profiling device and software (2). One of the most accurate numerical models for device profiling is achieved by the measurement of a large number of colors, which can be used to develop multi-dimensional LUTs with interpolation for any intermediate colors. The accuracy depends on color measurement. Lack of accuracy can lead to quantization effects. However, in practice, one must balance the time cost and the measurement. Thus, the selection of the number of measurement is a very challenging task in the design of LUTs. Nowadays, interpolation is widely used to decrease the number of measurements. Several interpolation methods, such as trilinear, prism, tetrahedral, etc., have been developed. Since there is more than one methods of interpolation, each with some errors, a situation arises where two CMMs, given identical input, can yield different results. Over the last few years, considerable progress have been made in instrument design and manufacture, which have led to more reliable instruments, stable readings and devices that are faster, lighter, and easier to use. Systematic errors, due to factors inherent in the manufacture of the instruments and the measuring situation, remain constant in time with respect to the selection and calibration of instruments and well-controlled measuring environments. However, the random errors, due to unpredictable variations during color measurement, are somehow instantaneous and unavoidable in the course of measurement, and can only be optimized by using the average of a number of repeated and consecutive measurements. The precision and uncertainty of…

Multispectral image acquisition typically results in a huge amount of data and often involves a c... more Multispectral image acquisition typically results in a huge amount of data and often involves a complicated mechanical setup in which a given number of interference filters is mounted on a filter wheel in front of a monochrome digital camera. Such constraints present serious problems for multispectral image capture over a wide area network. We solved these problems by using an electronically tunable spectral filter and a modular client-server software architecture. The client parts of the software are implemented in Java and allow for interactive operation of the multispectral camera across the Web.

We present the development at ENST of a multispectral imaging system. Various methods have been i... more We present the development at ENST of a multispectral imaging system. Various methods have been implemented and tested for the characterisation of spectral camera sensitivity, the optimal choice of filters and the reconstruction of the spectral reflectance of the imaged surfaces. We first used a 4k-linear array camera with a set of large band filters. We are now experimenting with

arXiv (Cornell University), May 24, 2019

This work covers multiple aspects of overt visual attention on 3D renders: measurement, projectio... more This work covers multiple aspects of overt visual attention on 3D renders: measurement, projection, visualization, and application to studying the influence of material appearance on looking behaviour. In the scope of this work, we ran an eye-tracking experiment in which the observers are presented with animations of rotating 3D objects. The objects were rendered to simulate different metallic appearance, particularly smooth (glossy), rough (matte), and coated gold. The eye-tracking results illustrate how material appearance itself influences the observer's attention, while all the other parameters remain unchanged. In order to make visualization of the attention maps more natural and also make the analysis more accurate, we develop a novel technique of projection of gaze fixations on the 3D surface of the figure itself, instead of the conventional 2D plane of the screen. The proposed methodology will be useful for further studies of attention and saliency in the computer graphics domain.

2019 IEEE/CVF International Conference on Computer Vision Workshop (ICCVW), 2019

Cracks on a painting is not a defect but an inimitable signature of an artwork which can be used ... more Cracks on a painting is not a defect but an inimitable signature of an artwork which can be used for origin examination, aging monitoring, damage identification, and even forgery detection. This work presents the development of a new methodology and corresponding toolbox for the extraction and characterization of information from an image of a craquelure pattern. The proposed approach processes craquelure network as a graph. The graph representation captures the network structure via mutual organization of junctions and fractures. Furthermore, it is invariant to any geometrical distortions. At the same time, our tool extracts the properties of each node and edge individually, which allows to characterize the pattern statistically. We illustrate benefits from the graph representation and statistical features individually using novel Graph Neural Network and hand-crafted descriptors correspondingly. However, we also show that the best performance is achieved when both techniques are merged into one framework. We perform experiments on the dataset for paintings origin classification and demonstrate that our approach outperforms existing techniques by a large margin.

ArXiv, 2019

Deep learning algorithms have demonstrated state-of-the-art performance in various tasks of image... more Deep learning algorithms have demonstrated state-of-the-art performance in various tasks of image restoration. This was made possible through the ability of CNNs to learn from large exemplar sets. However, the latter becomes an issue for hyperspectral image processing where datasets commonly consist of just a few images. In this work, we propose a new approach to denoising, inpainting, and super-resolution of hyperspectral image data using intrinsic properties of a CNN without any training. The performance of the given algorithm is shown to be comparable to the performance of trained networks, while its application is not restricted by the availability of training data. This work is an extension of original "deep prior" algorithm to HSI domain and 3D-convolutional networks.

In the last couple of decades, hyperspectral, multispectral, and multimodal (HMM) imaging has eme... more In the last couple of decades, hyperspectral, multispectral, and multimodal (HMM) imaging has emerged as an essential tool in various fields of science, medicine, and technology. Compared to integrated broad-band information as, e.g., present in RGB images, HMM imaging strives to acquire a multitude of specific narrow bands of the electromagnetic spectrum in order to solve specific detection or analysis tasks. HMM research is interested in studying light-matter interaction in a wide range of wavelengths from the high energy radiation down to Terahertz radiation (sub-millimeter waves). Furthermore, combining spectral data captured using different imaging modalities can unveil additional information of the scene that is not revealed solely by each of the individual imaging modalities. The workshop intended to connect researchers from different disciplines that involve HMM imaging and analysis. Even though there are very different approaches towards HMM imaging research and application...

Optics Express, 2021

A colour appearance model based on a uniform colour space is proposed. The proposed colour appear... more A colour appearance model based on a uniform colour space is proposed. The proposed colour appearance model, ZCAM, comprises of comparatively simple mathematical equations, and plausibly agrees with the psychophysical phenomenon of colour appearance perception. ZCAM consists of ten colour appearance attributes including brightness, lightness, colourfulness, chroma, hue angle, hue composition, saturation, vividness, blackness, and whiteness. Despite its relatively simpler mathematical structure, ZCAM performed at least similar to the CIE standard colour appearance model CIECAM02 and its revision, CAM16, in predicting a range of reliable experimental data.

2019 IEEE/CVF International Conference on Computer Vision Workshop (ICCVW), 2019

Deep learning algorithms have demonstrated state-ofthe-art performance in various tasks of image ... more Deep learning algorithms have demonstrated state-ofthe-art performance in various tasks of image restoration. This was made possible through the ability of CNNs to learn from large exemplar sets. However, the latter becomes an issue for hyperspectral image processing where datasets commonly consist of just a few images. In this work, we propose a new approach to denoising, inpainting, and superresolution of hyperspectral image data using intrinsic properties of a CNN without any training. The performance of the given algorithm is shown to be comparable to the performance of trained networks, while its application is not restricted by the availability of training data. This work is an extension of original "deep prior" algorithm to hyperspectral imaging domain and 3D-convolutional networks.

Color and Imaging Conference, 2018

Color Imaging: Device-Independent Color, Color Hardcopy, and Graphic Arts VI, 2000

Due to the increasing popularity and affordability of color imaging devices, color characterizati... more Due to the increasing popularity and affordability of color imaging devices, color characterization for these devices becomes an important subject. In other words, a set of color profile(s) needs to be generated for each device to transform the device dependent color space to a device independent one. This paper will concentrate on color characterization of scanners. Up to now, most scanner characterization has been optimized for photographic materials only. However, the color characterization inevitably depends on the media type and the printing mechanism that produced the target. In this work, a color scanner was characterized based on six different scan target types. For each target type, a color profile from RGB to CIELAB space was generated using polynomial regression method and a series of color conversions were conducted using color profiles optimized for itself as well as other target types. The accuracy of color transformation was evaluated. Results proved that to achieve the best color reproduction quality, it's necessary to characterize a color scanner based on specific scan target type. Considering the system level memory limitation, a compromise could be made without great quality degradation, using an average or unified profile for some target types. Nevertheless, a scanner should at least have two profiles: one for photo and the other for halftoning printing process.

SPIE Proceedings, 1999

To achieve high image quality throughout a digital imaging system, the first requirement is to en... more To achieve high image quality throughout a digital imaging system, the first requirement is to ensure the quality of the device that captures real-world physical images to digital images, for example a desktop scanner. Several factors have influence on this quality: optical resolution, bit depth, spectral sensitivities, and acquisition noise, to mention a few. In this study we focus on the colorimetric faculties of the scanner, that is, the scanner's ability to deliver quantitative device-independent digital information about the colors of the original document. We propose methods to convert from the scanner's device-dependent RGB color space to the standard device-independent color space sRGB. The methods have been evaluated using several different desktop scanners. Our results are very good: mean CIELAB (Delta) E*ab color errors as low as 1.4. We further discuss advantages and disadvantages of a digital color imaging system using the sRGB space for image exchange, compared to using other color architectures.

Lecture Notes in Computer Science, 2012

Based on previous research on super-resolution and colour correction by example, the present proj... more Based on previous research on super-resolution and colour correction by example, the present project employs such techniques towards a high-quality reclamation of lost art. The aim is to produce high-quality images of nowadays destroyed or missing paintings using the correspondence between similar artworks available in both low and high quality images. Several approaches for both super-resolution and colour correction techniques have been studied, implemented and tested to result to the most efficient and appropriate to be used. This project is an attempt that has never been done before and the successful reclamation of lost art initialises a new area of colour imaging applications in fine art. It reveals unlimited possibilities in the domain and establishes the potential of further attempts in this direction. iii

SPIE Proceedings, 1998

... Genevi eve Dardier, Jon Yngve Hardeberg and Hans Brettel ... Some manufacturers seem to take ... more ... Genevi eve Dardier, Jon Yngve Hardeberg and Hans Brettel ... Some manufacturers seem to take the necessary steps in order to signi cantly improve the quality of the pho-tographs by introducing a new type of image lters,1,2 using LCD technology, to correct areas on ...

A new method for the colorimetric characterization of a prin ter is proposed. It can also be appl... more A new method for the colorimetric characterization of a prin ter is proposed. It can also be applied to any other type of digital image reprodu ction device. The method is based on a computational geometry approach. It use s a 3D triangulation technique to build a tetrahedral partition of the pri nter color gamut volume and it generates a surrounding structure enclosing the defin ition domain. The characterization provides the inverse transformation from the device-independent color space CIELAB to the device-dependentcolor space CMY, takin g into account both colorimetric properties of the printer, and color gamut map ping.

Image reproduction suffers from several limitations in a color management system. In this paper, ... more Image reproduction suffers from several limitations in a color management system. In this paper, we have investigated artifacts resulting from the inherent characteristics of the color transformations by interpolation in three-dimensional look-up tables, and the unavoidable measurement noise of the color measurement done during device profiling. In our experiments, images were manipulated using three interpolation methods, and five levels of random noise. Psychophysical experiments were conducted to evaluate the quality of the reproduced images. Finally the experimental data were collected to analyze the color transformations, and test the performance of two color image difference metrics in this context. Introduction There has been increased demand to reproduce images using heterogeneous devices and media such as digital cameras, displays and printing systems. The employment of different color imaging devices results in a common problem that each device produces color differently. For example, to use the same values based on the device primaries, such as RGB for a display, would reproduce different colors by different printing systems. Hence, users are lacking color predictability and consistency to reproduce color images across different imaging media. This has been a driving force for the industry to develop technology to achieve successful cross-media color reproduction. Numerous attempts for the development of color management systems have been made to satisfy different image reproduction tasks from one medium to another. The most widely used systems are those based on the International Color Consortium (ICC) specifications. The ICC specification version 4 (1) provides definitions of color management architecture, profile format, and data structure. A typical ICC-based color management system consists of four basic components: profile, profile connection space (PCS), rendering intent, and color management module (CMM). A profile is a standard formatted file describing the device characterization, which defines the relationship between a device's control signals and the actual color that those signals produce. The ICC profile often employs multi-dimensional look- up tables (LUTs) to store the desired values. A process known as device characterization (or profiling) serves for this purpose, which provides a reliable way for color communications between media, and it is sufficient in simple applications with well-specified viewing conditions. The CMM is simply a color engine or processing engine, which is typically built into operating system, application or output device. The CMM performs all calculations needed to translate from the color space of one device to that of another. Although the ICC specifies the format of color profiles and to some extent the types of transforms that must be taken place to match colors between profiles, much of the process is left up to the imagination of the CMM creators. While it is difficult to specifically evaluate a vendor's commercial secrets, CMMs can certainly be evaluated based on the results they produce, both objectively by analysis of measurements and subjectively for pleasing contents. However, because different profiling applications will generate slightly different profiles from the same set of measurement data, the choice of CMM makes far less difference than the choice of profiling device and software (2). One of the most accurate numerical models for device profiling is achieved by the measurement of a large number of colors, which can be used to develop multi-dimensional LUTs with interpolation for any intermediate colors. The accuracy depends on color measurement. Lack of accuracy can lead to quantization effects. However, in practice, one must balance the time cost and the measurement. Thus, the selection of the number of measurement is a very challenging task in the design of LUTs. Nowadays, interpolation is widely used to decrease the number of measurements. Several interpolation methods, such as trilinear, prism, tetrahedral, etc., have been developed. Since there is more than one methods of interpolation, each with some errors, a situation arises where two CMMs, given identical input, can yield different results. Over the last few years, considerable progress have been made in instrument design and manufacture, which have led to more reliable instruments, stable readings and devices that are faster, lighter, and easier to use. Systematic errors, due to factors inherent in the manufacture of the instruments and the measuring situation, remain constant in time with respect to the selection and calibration of instruments and well-controlled measuring environments. However, the random errors, due to unpredictable variations during color measurement, are somehow instantaneous and unavoidable in the course of measurement, and can only be optimized by using the average of a number of repeated and consecutive measurements. The precision and uncertainty of…

Multispectral image acquisition typically results in a huge amount of data and often involves a c... more Multispectral image acquisition typically results in a huge amount of data and often involves a complicated mechanical setup in which a given number of interference filters is mounted on a filter wheel in front of a monochrome digital camera. Such constraints present serious problems for multispectral image capture over a wide area network. We solved these problems by using an electronically tunable spectral filter and a modular client-server software architecture. The client parts of the software are implemented in Java and allow for interactive operation of the multispectral camera across the Web.

We present the development at ENST of a multispectral imaging system. Various methods have been i... more We present the development at ENST of a multispectral imaging system. Various methods have been implemented and tested for the characterisation of spectral camera sensitivity, the optimal choice of filters and the reconstruction of the spectral reflectance of the imaged surfaces. We first used a 4k-linear array camera with a set of large band filters. We are now experimenting with