Writer Identification Using Differential Chain Codes and Grid Features

An approach of fast image mosaic is presented, which involves image matching and image intensity smoothing. Image matching includes two procedures, i.e. rough matching and fine matching. In rough matching, the overlapped regions of two adjacent images to be mosaicked are segmented to binary image at first. Then the binary images are filtered by open operation of mathematic morphologic method. In the binary image region of the reference image, feature template is searched and extracted on a given rule. Via XOR operation of the feature template and search region, some possible matching positions in the overlapped region of the other image are got. In the fine matching, the sequential similarity detection algorithm (SSDA) is adopted to perform matching computation in the small regions near the positions got in the rough matching, and then the relative position offsets in X-orientation and Y-orientation between the two adjacent images are got. Based on the result of the image matching, the two images are stitched. An approach of seam-line smoothing is adopted to adjust the intensity of the overlapped area. Simulation experimental results show that the approach greatly improves the operation speed, while the precision remains fine, so it can be applied in real-time mosaicking.

An approach of image mosaic based on binary image region matching is presented and applied in remote sensing serial image processing. Adopted the Nearest-neighbour interpolation, the original images are rectified on the basis of the values of external orientation elements of the camera installed on the unmanned aerial vehicle (UAV). Image matching includes two procedures, i.e. rough matching and fine matching. In rough matching, the area coverage of overlapped regions between two consecutive images to be stitched is determined approximately at first. Both of the overlapped regions of the two images are processed to binary image, and then the binary images are filtered by mathematic morphologic method. In the appointed binary image region of the former image, feature template is searched and extracted. Through the XOR operation of the feature template and search region, some possible matching positions in the overlapped region of the latter image are obtained. In the fine matching, the sequential similarity detection algorithm (SSDA) is adopted to perform matching computation in the some small regions near the positions extracted in the rough matching, and then the relative position offsets in X-orientation and Y-orientation between the two consecutive images are obtained. Based on the result of the image matching, the two images are stitched. Experimental results by simulation show that the approach presented in the paper greatly improves the operation speed, while the precision remains fine, than it can be applied in real-time mosaicking of serial images.

Chain coding of hardwritten line drawings for digital transmission is considered in this paper. A line drawing is considered to be composed of unconstrained two‐dimensional open and closed thin curves. Each curve is quantized in line‐following fashion into a sequence of discrete links, i.e., the chain‐code representation. Chain quantization based on grid‐intersect quantization (GIQ) and object boundary quantization (OBQ) is considered. Their performance in terms of the quantization distortion is quantitatively analyzed. To reduce the amount of data required to represent a line drawing, differential chain coding (DCC) is used to convert the chain code of a line drawing into binary data. The performance of DCC in terms of the number of bits required per unit length of curve is analyzed. The tradeoff between quantization distortion and coding efficiency for DCC is determined.

A new type of chain code, called multiring differential chain coding (MRDCC) is introduced. The coding rate is calculated and compared with standard chain coding (SCC), generalised chain coding (GCC), and differential chain coding (DCC) MRDCC offers the highest efficiency. Experimental observations are in good agreement with the theoretical results.

An adaptive fixed-length differential chain coding (FL-DCC) scheme is introduced for transmission of line graphics. Compared to differential chain coding (DCC), which uses vectors with different lengths, FL-DCC has a lower coding rate and a more uniform data syntax while maintaining a comparable subjective graphical quality.

A contour-based approach to binary shape coding using a multiple grid chain code

We propose a new fingerprint image compression scheme based on the hybrid model of an image. Our scheme uses the essential steps of a typical automated fingerprint identification system (AFIS) such as enhancement, binarization and thinning to encode fingerprint images. The decoding process is based on reconstructing a hybrid surface by using the gray values on ridges and valleys. In this compression scheme, the ridge skeleton is coded efficiently by using differential chain codes. The valley skeleton is derived from the ridge skeleton and the gray values along the ridge and valley skeletons are encoded using the discrete cosine transform. The error between the original and the replica is also encoded to increase the quality. One advantage of our approach is that original features such as end points and bifurcation points can be extracted directly from compressed image even for a very high compression ratio. Another advantage is that the proposed scheme can be integrated to a typical AFIS easily. The algorithm has been applied to various fingerprint images, and high compression ratios like 63:1 have been obtained. A comparison to wavelet/scalar quantization (WSQ) has been also made.

In differential chain coding (DCC), the bandwidth efficient relative vector is sensitive to channel errors, resulting in error propagation and corruption of the decoding process. A robust differential chain coding scheme is presented for stopping error propagation in the relative vector without increasing its bandwidth or using an additional forward error correction code. In this scheme, the relative vector is combined with the previous absolute vector to form a new error-resistant channel codeword. Experiments showed that, compared to DCC, the new scheme notably improves subjective quality when subject to transmission errors.

This paper proposes a novel method to text independent writer identification from online handwriting. The main contributions of our method include two parts: shape primitive representation and hierarchical structure. Both shape primitive's features are developed to represent the robust and distinctive characteristics of handwriting in two hierarchies. In first hierarchy, the shape primitives probability distribution function (SPPDF)is defined as the static features, to characterize orientation information of writing style. For each shape primitive, the statistics of pressure is defined as the dynamic shape primitives probability distribution function (DSPPDF) and the second hierarchy we build Gaussian model in dynamic attributes (DA) according to curvature of shape primitives. Experiments were conducted on the NLPR handwriting database collected from 242 persons. The results show that the new method achieves high accuracy, fast speed and low requirement of the amount of characters in handwriting samples. We achieve a writer identification rate of 91.5% with datasets in Chinese text and 93.6% in English text.

In a criminal investigation, a writer identification is frequently performed to find out what kind of writer a certain letter was written. In comparing the similarity between the handwriting of both documents, it is necessary to extract a person's characteristic writing style. Although the text-dependent methods have been studied show high discrimination performance for a writer identification, the situation assumed in the actual appraisal that the same character class does not exist is not supported. In this research, we propose a writer identification method taking into account the actual appraisal situation. We define the handwriting features without dependence on character class as writing and construct the AutoEncoder with the condition of character class to extract personal writing style from a single character sample. In the latent space trained by the conditional AutoEncoder, similar personal writing styles are mapped on to neighboring points independent of the character class. At the time of writer identification, the similarity between the handwriting feature of unknown writer and reference writer in the latent space is evaluated. In order to confirm the effectiveness of the proposed method, we conducted a writer identification experiments using ETL-1 Character Database and NIST Special Database 19 2^nd Edition. As a result, it is indicated that it is possible to extract personal writing style, which is an effective feature for writer identification, even under conditions close to the practical situation.

Recently, there has been growing interest in the study on cohesion of texts, particularly in the aspect of anaphoric reference. The problem under discussion is within the scope of similar and different usages of anaphors in English and Chinese. Although much work has been done, more studies need to be conducted to ascertain the convincingness of the theories. This essay will demonstrate language materials in real situation in English text and its Chinese translations to prove, explain and enrich the theories about the distinction of English and Chinese anaphora. The thesis for this essay is the contrastive study of anaphora between English and Chinese texts that is the dissimilarities of anaphora in these two language texts and discuss the reasons that caused these differences. And for arguments, four cases of Han’s Andersen’s fairy tales and their Chinese translations and one example from New Oxford Dictionary are selected for the purpose. For methodology, quality analysis is employed. But in general, it is to compare and to contrast in the light of Halliday’s theory on cohesion. The way adopted is to compare broad wise, which means to list similarities and dissimilarities of things needed to be contrast -- anaphora in English and Chinese texts, and then conducts further analysis on them with the theoretical framework. In aspect of detailed analysis, the author adopts the way of illustration, combining language phenomenon listed above with arguments.

Deep Convolutional Neural Networks (CNN) have shown great success in supervised classification tasks such as character classification or dating. Deep learning methods typically need a lot of annotated training data, which is not available in many scenarios. In these cases, traditional methods are often better than or equivalent to deep learning methods. In this paper, we propose a simple, yet effective, way to learn CNN activation features in an unsupervised manner. Therefore, we train a deep residual network using surrogate classes. The surrogate classes are created by clustering the training dataset, where each cluster index represents one surrogate class. The activations from the penultimate CNN layer serve as features for subsequent classification tasks. We evaluate the feature representations on two publicly available datasets. The focus lies on the ICDAR17 competition dataset on historical document writer identification (Historical-WI). We show that the activation features trained without supervision are superior to descriptors of state-of-the-art writer identification methods. Additionally, we achieve comparable results in the case of handwriting classification using the ICFHR16 competition dataset on historical Latin script types (CLaMM16).

This paper deals with identifying a writer from his/her offline handwriting. In a multilingual country where a writer can scribe in multiple scripts, writer identification becomes challenging when we have individual handwriting data in one script while we need to verify/identify a writer from handwriting in another script. In this paper such an issue is addressed with two scripts: English and Bengali. Here we model the task as a classification problem, where training data contains only Bengali handwritten samples and testing is performed on English handwritten texts. This work is based on the understanding that a writer has some inherent stroke characteristics that are independent of the script in which (s)he writes. In this work, some implicit structural and statistical features are extracted, and multiple classifiers are employed for writer identification. Many training sessions are run on a database of 100 writers and the performances are analyzed. We have obtained encouraging results on this database, which show the effectiveness of our method.

This paper proposes a method for texture feature extraction by integrating Gabor filters and independent component analysis (ICA) for Uyghur handwriting based writer identification. That is, the texture image is firstly filtered by a given bank of Gabor filters, and then higher dimensional feature vectors are constructed from the filtered texture images. Next, the dimensionality of these vectors is reduced by means of principal component analysis (PCA). Finally, the independent components in the resulting vectors with dimensionality reduced are analyzed and extracted by us. Experiments were performed using KNN-5 classifier to Uyghur handwriting samples from 55 different people and promising results of 92.5% correct identification rate were achieved.

Registration of pathological section images is an important part of three-dimensional reconstruction of sections. In this paper, a registration method was proposed to solve the problem of mismatching of pathological section images of pulmonary nodules. Firstly, rough matching is performed, and the feature points are extracted according to the scale-invariant feature transform (SIFT) algorithm. Then fast sparse coding (FSC) is used for fine matching to eliminate mismatched pairs. The algorithm presented in this paper is applied to the registration between sequence sections of pulmonary nodules. The experimental results show that the algorithm can effectively find more matching point pairs, accurately remove the false matching point pairs, and significantly improve the registration accuracy.