“…We report the F-measure in Table . 1, where our models outperforms others significantly. In our experiments, Zhang-Suen algorithm works better than other traditional methods, such as stroke correction (Pujari, Mitra, and Mishra 2014) and stroke continuity (Dong et al 2017). Distance-based methods are suitable for patterns with simple shapes and smooth contours, but fail to present comparable outputs in our task.…”
Section: Experiments On Skeleton Extraction Tasksmentioning
confidence: 80%
“…Most existing methods in character skeleton extraction focus on either local visual rules (Zhang and Suen 1984;Pujari, Mitra, and Mishra 2014;Dong et al 2017) or distance measurements (Zou and Yan 2001). These methods focus on low-level features in local regions, but when reading, humans turn to concern the skeletons of characters subconsciously and ignore the colors or widths of strokes.…”
Section: Related Work Deep Side Outputsmentioning
confidence: 99%
“…An improved ZhangSuen algorithm was designed for Odia characters, combining with stroke correction (Pujari, Mitra, and Mishra 2014). In (Dong et al 2017), stroke continuity detection serves as a preprocessing step for thinning. Recently, (Alghamdi and Teahan 2017) proposes a novel algorithm based on the boundary deletion with colour coding.…”
Section: Introductionmentioning
confidence: 99%
“…These methods are likely to yield unsatisfactory results when facing: (1) complex shapes, (2) variable stroke widths and (3) unsmooth edges. Particularly, the extracted lines are often distorted at the crosses or intersections of strokes (Dong et al 2017). These outcomes make stroke extraction and structural analysis difficult, while fully convolutional networks (FCNs) (Long, Shelhamer, and Darrell 2015) provide an pixel-to-pixel manner to solve these problems.…”
Structural analysis of handwritten characters relies heavily on robust skeletonization of strokes, which has not been solved well by previous thinning methods. This paper presents an effective fully convolutional network (FCN) to extract stroke skeletons for handwritten Chinese characters. We combine the holistically-nested architecture with regressive dense upsampling convolution (rDUC) and recently proposed hybrid dilated convolution (HDC) to generate pixel-level prediction for skeleton extraction. We evaluate our method on character images synthesized from the online handwritten dataset CASIA-OLHWDB and achieve higher accuracy of skeleton pixel detection than traditional thinning algorithms. We also conduct skeleton based character recognition experiments using convolutional neural network (CNN) classifiers on offline/online handwritten datasets, and obtained comparable accuracies with recognition on original character images. This implies the skeletonization loses little shape information.
“…We report the F-measure in Table . 1, where our models outperforms others significantly. In our experiments, Zhang-Suen algorithm works better than other traditional methods, such as stroke correction (Pujari, Mitra, and Mishra 2014) and stroke continuity (Dong et al 2017). Distance-based methods are suitable for patterns with simple shapes and smooth contours, but fail to present comparable outputs in our task.…”
Section: Experiments On Skeleton Extraction Tasksmentioning
confidence: 80%
“…Most existing methods in character skeleton extraction focus on either local visual rules (Zhang and Suen 1984;Pujari, Mitra, and Mishra 2014;Dong et al 2017) or distance measurements (Zou and Yan 2001). These methods focus on low-level features in local regions, but when reading, humans turn to concern the skeletons of characters subconsciously and ignore the colors or widths of strokes.…”
Section: Related Work Deep Side Outputsmentioning
confidence: 99%
“…An improved ZhangSuen algorithm was designed for Odia characters, combining with stroke correction (Pujari, Mitra, and Mishra 2014). In (Dong et al 2017), stroke continuity detection serves as a preprocessing step for thinning. Recently, (Alghamdi and Teahan 2017) proposes a novel algorithm based on the boundary deletion with colour coding.…”
Section: Introductionmentioning
confidence: 99%
“…These methods are likely to yield unsatisfactory results when facing: (1) complex shapes, (2) variable stroke widths and (3) unsmooth edges. Particularly, the extracted lines are often distorted at the crosses or intersections of strokes (Dong et al 2017). These outcomes make stroke extraction and structural analysis difficult, while fully convolutional networks (FCNs) (Long, Shelhamer, and Darrell 2015) provide an pixel-to-pixel manner to solve these problems.…”
Structural analysis of handwritten characters relies heavily on robust skeletonization of strokes, which has not been solved well by previous thinning methods. This paper presents an effective fully convolutional network (FCN) to extract stroke skeletons for handwritten Chinese characters. We combine the holistically-nested architecture with regressive dense upsampling convolution (rDUC) and recently proposed hybrid dilated convolution (HDC) to generate pixel-level prediction for skeleton extraction. We evaluate our method on character images synthesized from the online handwritten dataset CASIA-OLHWDB and achieve higher accuracy of skeleton pixel detection than traditional thinning algorithms. We also conduct skeleton based character recognition experiments using convolutional neural network (CNN) classifiers on offline/online handwritten datasets, and obtained comparable accuracies with recognition on original character images. This implies the skeletonization loses little shape information.
“…These methods are likely to yield unsatisfactory results when facing: (1) complex shapes, (2) variable stroke widths and (3) unsmooth edges. Particularly, the extracted lines are often distorted at the crosses or intersections of strokes [30] . Recently, FCN based skeletonization has been proven to outperform the above methods remarkably [31] , but for training FCN, it is infeasible to label skeleton pixels for millions of offline handwritten samples [32] .…”
Section: Skeletonization Of Handwritten Charactersmentioning
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.