Fig 1 - uploaded by Khurram Khurshid
Content may be subject to copyright.
A multispectral image represented as an image stack, spectral response at the spatial location (x,y), an RGB image and a grayscale image rendered from the image stack.
Source publication
Multispectral imaging allows for analysis of images in multiple spectral bands. Over the past three decades, airborne and satellite multispectral imaging have been the focus of extensive research in remote sensing. In the recent years, ground based multispectral imaging has gained immense interest in the areas of computer vision, medical imaging, a...
Contexts in source publication
Context 1
... cameras can capture reflectance, luminescence, photoluminescence and transmission in various parts of the electromagnetic spectrum such as ultraviolet, visible, infrared and thermal range. A multispectral image is an image stack with each image captured at a narrow spectral band and thus it contains spatial as well as spectral information as shown in Figure 1 using pseudo-colors. The spatial information relates to the geometric relationship of the image pixels to each other while the spectral information relates to the variations within image pixels as a function of wavelength. ...
Context 2
... multispectral image has two spatial dimensions (Sx and Sy) and one spectral dimension (S). The spectral response at the spatial location (x,y), an RGB image and a grayscale image rendered from the image stack are also shown in Figure 1. Spectral response of a single pixel in a multispectral image provides information about its constituents and surface of the material. ...
Context 3
... cameras can capture reflectance, luminescence, photoluminescence and transmission in various parts of the electromagnetic spectrum such as ultraviolet, visible, infrared and thermal range. A multispectral image is an image stack with each image captured at a narrow spectral band and thus it contains spatial as well as spectral information as shown in Figure 1 using pseudo-colors. The spatial information relates to the geometric relationship of the image pixels to each other while the spectral information relates to the variations within image pixels as a function of wavelength. A multispectral image has two spatial dimensions (Sx and Sy) and one spectral dimension (S). The spectral response at the spatial location (x,y), an RGB image and a grayscale image rendered from the image stack are also shown in Figure 1. Spectral response of a single pixel in a multispectral image provides information about its constituents and surface of the material. Multispectral imaging was initially developed for remote sensing purposes but it has found numerous applications in ground based imaging in the recent years, including agricultural and water resources control [11][12], military defense, art conservation and archeology [13][14], medical diagnosis [15][16], analyses of crime scene details [17][18], document imaging [19], forensic medicine [20], food quality control [21][22] and mineralogical mapping of earth surface [23]. The broad spectral information provided by multispectral images combined with signal processing allows for identification of the underlying material in images. Materials which are clearly visible in a specific band and obscured in the other bands are reflective within the same band and thus easily detectable by the multispectral imaging system [24]. Selection of appropriate bands in the multispectral data of historic texts and manuscripts allows for identification of inks and pigments for dating of manuscripts [25] and recovery of erased and overwritten scripts [26]. Over the past few years, various methods based on multispectral image analysis have been proposed for automated forgery detection in documents [27][28][29]. The use of multispectral image analysis have had a significant impact on forgery detection as compared to the traditional techniques. A multispectral imaging system proposed by Brauns et al. [30] detects forgery in a non-destructive manner in potentially fake documents using an interferometer which uses different moving parts for frequency tuning and subsequently moderates or slows down the acquisition procedure. This work verifies the concept to distinguish between writing inks. Subjective outcomes on a very small database demonstrated that the spectra of ink can be isolated into various classes in an unsupervised way. An advanced and more complex multispectral imaging framework for examination of ancient documents was created at the National Archives of Netherlands [31] which provided high spatial as well as high spectral resolution over a broad spectral range (near UV to near IR). The system was very robust and effective but its acquisition time is very long [32]. Texture [33] and ink-deposition traces [34] have also been employed for writing analysis in document images. Different inks were successfully classified after obliteration of the text in a document ...
Context 4
... cameras can capture reflectance, luminescence, photoluminescence and transmission in various parts of the electromagnetic spectrum such as ultraviolet, visible, infrared and thermal range. A multispectral image is an image stack with each image captured at a narrow spectral band and thus it contains spatial as well as spectral information as shown in Figure 1 using pseudo-colors. The spatial information relates to the geometric relationship of the image pixels to each other while the spectral information relates to the variations within image pixels as a function of wavelength. A multispectral image has two spatial dimensions (Sx and Sy) and one spectral dimension (S). The spectral response at the spatial location (x,y), an RGB image and a grayscale image rendered from the image stack are also shown in Figure 1. Spectral response of a single pixel in a multispectral image provides information about its constituents and surface of the material. Multispectral imaging was initially developed for remote sensing purposes but it has found numerous applications in ground based imaging in the recent years, including agricultural and water resources control [11][12], military defense, art conservation and archeology [13][14], medical diagnosis [15][16], analyses of crime scene details [17][18], document imaging [19], forensic medicine [20], food quality control [21][22] and mineralogical mapping of earth surface [23]. The broad spectral information provided by multispectral images combined with signal processing allows for identification of the underlying material in images. Materials which are clearly visible in a specific band and obscured in the other bands are reflective within the same band and thus easily detectable by the multispectral imaging system [24]. Selection of appropriate bands in the multispectral data of historic texts and manuscripts allows for identification of inks and pigments for dating of manuscripts [25] and recovery of erased and overwritten scripts [26]. Over the past few years, various methods based on multispectral image analysis have been proposed for automated forgery detection in documents [27][28][29]. The use of multispectral image analysis have had a significant impact on forgery detection as compared to the traditional techniques. A multispectral imaging system proposed by Brauns et al. [30] detects forgery in a non-destructive manner in potentially fake documents using an interferometer which uses different moving parts for frequency tuning and subsequently moderates or slows down the acquisition procedure. This work verifies the concept to distinguish between writing inks. Subjective outcomes on a very small database demonstrated that the spectra of ink can be isolated into various classes in an unsupervised way. An advanced and more complex multispectral imaging framework for examination of ancient documents was created at the National Archives of Netherlands [31] which provided high spatial as well as high spectral resolution over a broad spectral range (near UV to near IR). The system was very robust and effective but its acquisition time is very long [32]. Texture [33] and ink-deposition traces [34] have also been employed for writing analysis in document images. Different inks were successfully classified after obliteration of the text in a document ...
Citations
... The review by Khan et al. [15] underscores the extensive applications of hyperspectral imaging and illuminates the potential of an automatic forgery detection system integrating HSI with deep learning. Further advancing the field, Khan [21] implemented Fuzzy C-Means Clustering (FCM) to segregate the spectral responses of ink pixels into distinct clusters for ink combinations with varying mixing ratios, selecting spectral bands with the highest information content to refine the results. Beyond traditional pattern recognition methods, deep learning techniques have increasingly been integrated into HSI document ink mismatch detection. ...
... Despite the reduced number of replicates in each database, the present technique achieved values not much lower than those obtained in other studies with other techniques such as Micro-ATR-FTIR combined with chemometrics in which the discrimination power did not exceed 86.67% [6] and 89.99% [7], or in studies where multispectral imaging coupled with FCM was used and achieved discrimination percentages of 76% [8] and 85.7% [9]. ...
The requirement for the preservation of evidence in court continues to pose a challenge to forensic document analysis, where the answer to the discrimination of writing utensils is recurrent in requested questions. This study aims to contribute to the identification of the blue and green inks of handwriting instruments using Raman spectroscopy. Based on the Raman spectra obtained and after processing with Spectragryph software, a predictive model was built using KNIME software, where a discriminative power of 72.7% was obtained.
... HSI, on the other hand, is a reliable approach for non-destructive and non-contact forensic document inspection. HSI is a method that combines imaging and spectroscopy [10] [11]. In this method, each image is captured at a narrow band of the electromagnetic spectrum in order to collect precise spectral data. ...
... Several strategies for identifying ink mismatches based on HSI inspection have been proposed in the last decade. Fuzzy c-means clustering [10], k-means clustering [11], localized hyper spectral image analysis [20], and deep convolutional network [21] are among the approaches used. Jaleed et al. [10] proposed a multi spectral image analysis-based automatic ink mismatch detecting technique. ...
Forensic document examiners can determine the authenticity of questioned documents by analyzing the ink used to create them. If an ink mismatch is found, it could be a sign of scam, backdating, or forgery. In this research a Hyperspectral Images of iVision HHID dataset is used to detect number of possible inks used in document. By using Hyperspectral Images, it’s possible to detect ink mismatch in a given document. In this research unsupervised learning method K-means is used to detect number of inks. Approximate number of clusters are determined by Elbow and Silhouette method before implementation of K-means.
... The authors address the issue of disproportionate ink mismatch detection by identifying spectral signatures and proportions of inks. Automated forgery detection is another important area within forensic document analysis [7], which utilizes Fuzzy C-Means Clustering for ink mismatch detection in multispectral document images. This method outperforms existing techniques and shows promise for forensic analysis in various fields, including computational forensics. ...
In the field of document forensics, ink analysis plays a crucial role in determining the authenticity of legal and historic documents and detecting forgery. Visual examination alone is insufficient for distinguishing visually similar inks, necessitating the use of advanced scientific techniques. This paper proposes an ink analysis technique based on hyperspectral imaging, which enables the examination of documents in hundreds of narrowly spaced spectral bands, revealing hidden details. The main objective of this study is to identify the number of distinct inks used in a document. Three clustering algorithms, namely k-means, Agglomerative, and c-means, are employed to estimate the number of inks present. The methodology involves data extraction, ink pixel segmentation, and ink number determination. The results demonstrate the effectiveness of the proposed technique in identifying ink clusters and distinguishing between different inks. The analysis of a hyperspectral cube dataset reveals variations in spectral reflectance across different bands and distinct spectral responses among the 12 lines, indicating the presence of multiple inks. The clustering algorithms successfully identify ink clusters, with k-means clustering showing superior classification performance. These findings contribute to the development of reliable methodologies for ink analysis using hyperspectral imaging, enhancing the
... One limitation of this unmixing scheme is that HySime overestimates the number of inks present in the hyperspectral scene, thus necessitating manual discarding of non-coherent abundance maps. Ink mismatch detection in handwritten notes was performed in [17] using Fuzzy C-Means Clustering (FCM) to partition the spectral responses of ink pixels into distinct clusters. Experiments were performed on ink combinations with varying mixing ratios. ...
p>Document authentication is a critical part of forensic analysis, which ensures the veracity of a document’s origins. A significant challenge in this field is the detection of ink mismatches, especially in instances of disproportionate ink distribution. This issue is effectively handled with hyperspectral images of documents, whereas traditional imaging techniques struggle to distinguish between visually similar inks. To address this problem, we introduce a new approach that leverages hyperspectral unmixing for ink mismatch detection in unbalanced clusters. The proposed method identifies unique spectral characteristics of different inks and their respective proportions, thus facilitating significant ink distinction. The proposed approach utilizes Elbow estimation and Silhouette coefficient for number of inks estimation and performs color segmentation for all unique ink types by employing k-means clustering and Gaussian mixture models (GMMs), outperforming existing methods in this regard which rely on prior knowledge about the number of inks used in the document. For abundance estimation, a unique method of dimensional reduction of HSI and channel wise analysis is proposed. We evaluate our approach on the iVision Handwritten Hyperspectral Images Data set (iVision HHID), a comprehensive and rich dataset that surpasses the commonlyused UWA Writing Inks Hyperspectral Images (WIHSI) database in size and diversity. Our results, in comparison with state-of-theart methods, demonstrate the effectiveness of our approach in hyperspectral ink mismatch detection. This paper thus promotes the application of hyperspectral imaging for document analysis and encourages further exploration toward automated questioned document examination.</p
... One limitation of this unmixing scheme is that HySime overestimates the number of inks present in the hyperspectral scene, thus necessitating manual discarding of non-coherent abundance maps. Ink mismatch detection in handwritten notes was performed in [17] using Fuzzy C-Means Clustering (FCM) to partition the spectral responses of ink pixels into distinct clusters. Experiments were performed on ink combinations with varying mixing ratios. ...
p>Document authentication is a critical part of forensic analysis, which ensures the veracity of a document’s origins. A significant challenge in this field is the detection of ink mismatches, especially in instances of disproportionate ink distribution. This issue is effectively handled with hyperspectral images of documents, whereas traditional imaging techniques struggle to distinguish between visually similar inks. To address this problem, we introduce a new approach that leverages hyperspectral unmixing for ink mismatch detection in unbalanced clusters. The proposed method identifies unique spectral characteristics of different inks and their respective proportions, thus facilitating significant ink distinction. The proposed approach utilizes Elbow estimation and Silhouette coefficient for number of inks estimation and performs color segmentation for all unique ink types by employing k-means clustering and Gaussian mixture models (GMMs), outperforming existing methods in this regard which rely on prior knowledge about the number of inks used in the document. For abundance estimation, a unique method of dimensional reduction of HSI and channel wise analysis is proposed. We evaluate our approach on the iVision Handwritten Hyperspectral Images Data set (iVision HHID), a comprehensive and rich dataset that surpasses the commonlyused UWA Writing Inks Hyperspectral Images (WIHSI) database in size and diversity. Our results, in comparison with state-of-theart methods, demonstrate the effectiveness of our approach in hyperspectral ink mismatch detection. This paper thus promotes the application of hyperspectral imaging for document analysis and encourages further exploration toward automated questioned document examination.</p
... In 2018, J. Khan et al [9] applied Fuzzy C-Means on ink pixels (after removing background) to detect the number of clusters or inks present in a document image. They claimed 68% accuracy with this approach and improved it to 76% by applying feature selection, however, mixing of database is limited to 2 inks of same color by same subject and the method is required to be performed several times for real scenarios. ...
Writing was not a common-man’s job in old times and resources
involved in writing process were not abundant. Since the writing
surfaces remained limited to leaves, cloth, papyrus, wooden-sheets and limited-paper, over-writing or multiple writings on the available manuscript was a common practice. Analysis of such manuscripts require a distinction between various additions made over time.
... To classify different inks in a document, authors employed local thresholding to separate text and background pixels. They then applied fuzzy c-means clustering for the classification process, and for optimal results, feature selection techniques were utilized [19]. ...
p>Hyperspectral images are known to contain an extensive range of bands that offer a wealth of information that tri-spectral images cannot match. With an increased number of bands, hyperspectral images provide an enhanced level of reflectance and absorption values, making it possible to obtain more precise details at each point in the image. The HSI field continues to evolve and has already proven useful in various applications such as age prediction, handwritten OCR, word segmentation, ecological and hydrological science, forensics, and much more. This study involves showing the total number of bands of the hyperspectral image along with the starting and ending wavelength, displaying the 1st, 30th, 60th and last band of the HSI image, plotting the spectral responses of foreground pixels and applying k-means clustering to determine the number of inks present in the hyperspectral image document. Additionally, color-labeling techniques were utilized to identify the various colors of ink in the document.</p
... To classify different inks in a document, authors employed local thresholding to separate text and background pixels. They then applied fuzzy c-means clustering for the classification process, and for optimal results, feature selection techniques were utilized [19]. ...
p>Hyperspectral images are known to contain an extensive range of bands that offer a wealth of information that tri-spectral images cannot match. With an increased number of bands, hyperspectral images provide an enhanced level of reflectance and absorption values, making it possible to obtain more precise details at each point in the image. The HSI field continues to evolve and has already proven useful in various applications such as age prediction, handwritten OCR, word segmentation, ecological and hydrological science, forensics, and much more. This study involves showing the total number of bands of the hyperspectral image along with the starting and ending wavelength, displaying the 1st, 30th, 60th and last band of the HSI image, plotting the spectral responses of foreground pixels and applying k-means clustering to determine the number of inks present in the hyperspectral image document. Additionally, color-labeling techniques were utilized to identify the various colors of ink in the document.</p
... Tasks like OCR [4], [6], [7], [19], Text/non-Text segmentation [20], [21], Table Detection (TD) [22]- [24], Word Spotting (WS) [25]- [27], Document Categorization (DC), Script Identification (SI) [2], [3], [28], [29], Writer Identification (WI) [30]- [33] and Post-processing have not been considered. There are some specific topics which have been researched with respect to the DAR, such as, Forgery detection [34], Image quality assessment [35], Strike-off removal in handwritten scripts [36], [37] etc. ...
p>The journey of research for Document Analysis and Recognition (DAR) started with the problem of automatic character recognition. Today, it has covered a vast span of recognition tasks such as text recognition, script identification, writer identification, word spotting, signature verification etc., in scripts like Roman, Arabic, Chinese, Japanese, Brahmi etc. Extensive advancements in deep learning techniques have achieved state-of-the-art results for various DAR tasks. This work explores the challenges from different perspectives and reviews the techniques for online/offline and printed/handwritten DAR tasks. We examine the research works with the view of script-related challenges. Various datasets for DAR are categorized into historic, printed and handwritten datasets. We present a comprehensive survey of challenges, techniques, datasets, evaluation metrics, script-related perspectives and potential future directions for DAR.</p
