Text Regions Research Articles

Dental Image Segmentation and Classification Using Inception Resnetv2

The automated process for dental caries detection draws increasing attention with the technological innovation in machine learning methods. This is a core issue in dental diseases especially in the detection of caries as it leads to serious health ailments. This paper takes an effort to adequately segment and identify dental diseases. There are four main steps. The preprocessing technique uses binary histogram equalization which increases the texture region visibility for the caries detected on dental images. The novel technique of segmentation with Curvilinear Semantic Deep Convolutional Neural Network (CSDCNN) is proposed in this paper . The segmentation is followed by the proposed Inception resnetV2, which acts as the classification technique to determine the caries in dental images. The proposed segmentation algorithm is used to determine a dental degree of membership. The inception is brought out with different scales of information, which relates to various input images as data. An examination of the x-ray images will detect the impact of illness on a tooth. Particularly for the segmentation and classification mission, we deemed four diseases: dental caries, periapical infection, periodontal, and pericoronal diseases. Based on the number of input functional parameters, the Inception resnetV2 classifies different image categories effectively. The proposed Inception resnetV2 has become the most effective tool in machine learning to solve problems like image classification with a high order of accuracy. The average accuracy of the device proposed is 94.51%. This provides higher classification accuracy when compared to other existing methods.

Full-Reference Quality Assessment for Screen Content Images Based on the Concept of Global-Guidance and Local-Adjustment

Benefiting from the development of multimedia communication terminals, the visual content presented to people on mobile devices is no longer a single form, but contains text, natural images, and other computer-generated graphics, which is called screen content images (SCIs). Inspired by the different visual stimuli that text and images bring to human eyes and the concept of global-guidance and local-adjustment, we design a novel full-reference image quality assessment (IQA) model using the structural features of the text, the perceptual features of pictures, and a score integration model (SPSIM) to evaluate SCIs quality. Firstly, we split the SCIs into textual and pictorial regions through a fully convolutional network (FCN) to conduct separate analyses. For textual regions, we take advantage of narrow edge extensions and high edge steps as structural features to compute the textual score. For pictorial regions, we extract the just noticeable difference (JND) features, which measure the human eye's ability to detect distortion as perceptual features to calculate the pictorial score. Finally, an innovative score integration method based on the global-guidance and local-adjustment is designed to better analyze the relationship between the above regional scores and the whole global SCIs score. Abundant experiments in SCIs databases have shown that the SPSIM model can achieve better consistency with the human eyes system (HVS) in predicting the visual quality of SCIs.

Sequential alignment attention model for scene text recognition

Scene text recognition has been a hot research topic in computer vision due to its various applications. The state-of-the-art solutions usually depend on the attention-based encoder-decoder framework that learns the mapping between input images and output sequences in a purely data-driven way. Unfortunately, there often exists severe misalignment between feature areas and text labels in real-world scenarios. To address this problem, this paper proposes a sequential alignment attention model to enhance the alignment between input images and output character sequences. In this model, an attention gated recurrent unit (AGRU) is first devised to distinguish the text and background regions, and further extract the localized features focusing on sequential text regions. Furthermore, CTC guided decoding strategy is integrated into the popular attention-based decoder, which not only helps to boost the convergence of the training but also enhances the well-aligned sequence recognition. Extensive experiments on various benchmarks, including the IIIT5k, SVT, and ICDAR datasets, show that our method substantially outperforms the state-of-the-art methods.

Adaptive Directional Cubic Convolution for Integrated Circuit Chip Defect Image Interpolation

An adaptive directional cubic convolution interpolation method for integrated circuit (IC) chip defect images is proposed in this paper, to meet the challenge of preserving edge and texture information. In the proposed method, Otsu thresholding technique is employed to distinguish strong edge pixels from weak ones and texture regions, and estimate the direction of strong edges, adaptively. Boundary pixels are pre-interpolated using the original bicubic interpolation method to help improve the interpolation accuracy of the interior pixels. The experimental results of both classic test images and IC chip defect images demonstrate that the proposed method outperforms the competing methods with better edge and texture preservation, interpolation quality, more natural visual effect of the interpolated images and reasonable computational time. The proposed method can provide high quality IC chip images for defect detection and has been successfully applied on practical vision inspection for IC chips

Mproved Stereo Matching Algorithm based on Census Transform and Dynamic Histogram Cost Computation

Stereo matching is a significant subject in the stereo vision algorithm. Traditional taxonomy composition consists of several issues in the stereo correspondences process such as radiometric distortion, discontinuity, and low accuracy at the low texture regions. This new taxonomy improves the local method of stereo matching algorithm based on the dynamic cost computation for disparity map measurement. This method utilised modified dynamic cost computation in the matching cost stage. A modified Census Transform with dynamic histogram is used to provide the cost volume. An adaptive bilateral filtering is applied to retain the image depth and edge information in the cost aggregation stage. A Winner Takes All (WTA) optimisation is applied in the disparity selection and a left-right check with an adaptive bilateral median filtering are employed for final refinement. Based on the dataset of standard Middlebury, the taxonomy has better accuracy and outperformed several other state-ofthe-art algorithms. Keywords—Stereo matching, disparity map, dynamic cost, census transform, local method

Conceptual text region network: Cognition-inspired accurate scene text detection

Segmentation-based methods are widely used for scene text detection due to their superiority in describing arbitrary-shaped text instances. However, two major problems still exist: (1) current label generation techniques are mostly empirical and lack theoretical support, discouraging elaborate label design; and (2) as a result, most methods rely heavily on text kernel segmentation which is unstable and requires deliberate tuning. To address these challenges, we propose a human cognition-inspired framework, termed, Conceptual Text Region Network (CTRNet). The framework utilizes Conceptual Text Regions (CTRs), which is a class of cognition-based tools inheriting good mathematical properties, allowing for sophisticated label design. Another component of CTRNet is an inference pipeline that, with the help of CTRs, completely omits the need for text kernel segmentation. Compared with previous segmentation-based methods, our approach is not only more interpretable but also more accurate. Experiment results show that CTRNet achieves state-of-the-art performance on benchmark CTW1500, Total-Text, MSRA-TD500, and ICDAR 2015 datasets, yielding performance gains of up to 2.0%. Notably, to the best of our knowledge, CTRNet is among the first detection models to achieve F-measures higher than 85.0% on all four of the benchmarks, demonstrating remarkable consistency and stability.

Multi-focus Source Images Reconstruction based on Adaptive Regional Data Hiding

Preserving the information of multi-focus source images was neglected in previous image fusion schemes since source image data discarding will happen during the fusion process. Data hiding technology can utilize the redundant bits to embed essential secret data. Inspired from that we proposed a multi-focus image reconstruction algorithm using adaptive regional data hiding. Integrating the Human Visual System (HVS) into the reconstruction module, we further proposed a visual grayscale information entropy operator, which is implemented to segment fused images into texture and flat regions for adaptive data hiding after unfocused region data compression. Our method achieves excellent performances in reconstruction Peak signal-to-noise Ratio (PSNR) above 43dB and maintains the satisfying visual effect of the fused images.

Research on Remote Sensing Image Matching with Special Texture Background

The purpose of image registration is to find the symmetry between the reference image and the image to be registered. In order to improve the registration effect of unmanned aerial vehicle (UAV) remote sensing imagery with a special texture background, this paper proposes an improved scale-invariant feature transform (SIFT) algorithm by combining image color and exposure information based on adaptive quantization strategy (AQCE-SIFT). By using the color and exposure information of the image, this method can enhance the contrast between the textures of the image with a special texture background, which allows easier feature extraction. The algorithm descriptor was constructed through an adaptive quantization strategy, so that remote sensing images with large geometric distortion or affine changes have a higher correct matching rate during registration. The experimental results showed that the AQCE-SIFT algorithm proposed in this paper was more reasonable in the distribution of the extracted feature points compared with the traditional SIFT algorithm. In the case of 0 degree, 30 degree, and 60 degree image geometric distortion, when the remote sensing image had a texture scarcity region, the number of matching points increased by 21.3%, 45.5%, and 28.6%, respectively and the correct matching rate increased by 0%, 6.0%, and 52.4%, respectively. When the remote sensing image had a large number of similar repetitive regions of texture, the number of matching points increased by 30.4%, 30.9%, and −11.1%, respectively and the correct matching rate increased by 1.2%, 0.8%, and 20.8% respectively. When processing remote sensing images with special texture backgrounds, the AQCE-SIFT algorithm also has more advantages than the existing common algorithms such as color SIFT (CSIFT), gradient location and orientation histogram (GLOH), and speeded-up robust features (SURF) in searching for the symmetry of features between images.

An improved graph cut algorithm in stereo matching

In stereo matching tasks, the matching effect is often very poor when the texture of the region is weak or repeated. To solve this problem, an improved Graph Cut stereo matching algorithm based on Census transform is proposed. The Hamming distance of the corresponding pixels in the left and right images after Census transform is introduced as the similarity measure in the data term of the energy function. In this way, the dependence on the pixel value is reduced. The stereo matching experiments are implemented on the standard images of Middlebury stereo benchmark and the real scene images, and it demonstrates that our algorithm is robust and can obtain better performance in weak texture or repeated texture region.

Weather Prediction and Smart Notification System

In the recent period, many real-world applications and institutions generates a huge amount of data which is unstructured i.e., in the form of images containing data, receipts, invoices, forms, statements, contracts etc. This rich and detailed information presented in the text is of great significance in computer vision-based applications (driverless cars, assisting blind and visually impaired people, detecting labels and packages, automatic number plate recognition etc.). Recently, there has been a hike in the efforts, researches and progresses being done in this domain for its significance in data analysis and computer vision. Here has been a diversity of challenges in unstructured data like image sensor noise, different viewing angles, blur, lighting conditions, resolution, and non-planar object. Our objective for taking up this topic for research are i) to detect and recognize the text from the data ii) to handle diversity and variability of text in natural scene iii) to explore various datasets iv) to deal with various issues occurring in scene text detection. To tackle this problem, we propose a robust scene text detection and recognition method with adaptive text region representation using deep learning model open CV with EAST algorithm as detection pipeline and tesseract. The recurrent neural network-based adaptive text region representation is proposed for text region refinement, where a pair of boundary points are predicted each time step until no new points are found. In this way, text regions in an image are detected and represented with the adaptive number of boundary points.

Differential diagnosis of left ventricular hypertrophy with cardiac magnetic resonance imaging texture feature analysis

Abstract Funding Acknowledgements Type of funding sources: None. Background Hypertrophic cardiomyopathy (HCM), hypertensive cardiomyopathy (HTCM), and cardiac amyloidosis (CA) are characterized by left ventricular (LV) hypertrophy. CMR with late gadolinium enhancement is used in the differential diagnosis. Radiomics aims to extract features from medical images and quantify their patterns to produce quantitative biomarkers. Our aim was to develop radiomics-derived biomarkers capable of differentiating between HCM, HTCM, CA and healthy controls using CMR cine images. Methods We included 103 patients (42 HCM, 51 HTCM, 10 CA; 31 females, age 65.0 ± 12.8 years) and 50 healthy subjects (29 females, age 52.5 ± 18.0 years). The study was done on a 1.5 T CMR scanner (Sonata; Siemens, Erlangen, Germany) and all the subjects underwent a conventional protocol including cine sequences acquired in the 2-, 4- and 3-chamber views as well as in a short axis stack. The most basal slice of the short axis stack immediately distal to the LV outflow tract was used for texture feature analysis and regions of interest (ROI) were defined by segmenting the LV myocardium and dividing it into 6 segments. For image analysis a radiomic approach was used. Five different quantizations were obtained in order to find the optimal one and 43 texture features were extracted by quantization and segment. Feature selection (FS) methods were studied to see the optimal number of textures, and 4 predictive models were evaluated including k-Nearest Neighbour (KNN) and Support Vector Machine with a linear kernel (SVML) models. Performance of each model was evaluated with the area under the ROC curve (AUC). Results The AUCs of each of the models for the different texture features were analysed, from 1 to 43, according to the ranking obtained and for 5 different number of grey levels. The maximum average AUC, 1.0, was obtained with KNN model trained to classify CA and control subjects. More interestingly, 0.89 average AUC was achieved with SVML model and 1 single feature for HCM vs HTCM vs CA study. Table 1 shows the best average AUCs (mean ± SD) obtained for each study, number of subjects (N), number of grey levels used in each case, number of optimal features and the classification model. Conclusions Texture analysis allows for the correct differential diagnosis of LVH. Best results (AUC of 89%), were achieved using the SVML model in septal segments. Texture feature analysis applied to CMR cine sequences is a promising advance that might allow for the use of contrast-free protocols for the differential diagnosis of LVH.

Adaptive preprocessing of character recognition image based on neural network

Words in natural scenes contain abundant information. Automatic acquisition of text information in images can help people understand images more effectively and further process images such as storage, compression and retrieval. Pattern recognition technology is to use machines to simulate people’s various recognition abilities. At present, it mainly simulates people’s visual and auditory abilities. Based on BP neural network in deep learning, a text positioning BP neural network is proposed, which automatically extracts the text features in the image and avoids the defect of using manual design features This paper studies the preprocessing algorithm of Chinese character recognition in natural scenes, including binarization of text regions, extraction of character colors, clustering analysis of colors, segmentation of characters, etc., and designs the rules for merging text regions after positioning. Experimental results show that the algorithm can solve the problem of insufficient recognition ability for text images and long-sequence text images with complex background noise between characters when the background and text gray level are similar.

Improved Face Recognition With Fractal-Based Texture Analysis

Fractals are useful to uniquely represent texture in the human face, which serves as an equivalent of human vision. FaceNet, calculating face descriptors of a person, has been observed to perform with setbacks when several factors of occlusion are present. This paper proposes a new methodology that exploits the self-similar patterns in a person's face to highlight and enhance regions of high texture in a facial image. The system maps the original image into a representation in the pre-processing stage of computer vision. This representation when fed as an input to the FaceNet CNN optimizes the face embedding generated. An SVM classifier separates the hard positive examples from the hard negative examples during classification. The model is trained using YouTube Faces DB as primary dataset and for validation; a custom dataset is designed to verify a person's identity despite the presence of secondary factors such as expressions and forgery. The proposed model attained an overall accuracy of 96.73% with the YouTube Faces DB, and a notable reduction in the false positive rates is observed.

HYBRID TEXTURE IDENTIFICATION METHOD

The importance of the modeling mode in systems of computer visual pattern recognition is shown. The purpose of the mode is to determine the types of textures that are present on the images processed in intelligent diagnostic systems. Images processed in technical diagnostic systems contain texture regions, which can be represented by different types of textures -spectral, statistical and spectral-statistical. Texture identification methods, such as, statistical, spectral, expert, multifractal, which are used to identify and analyze texture images, have been analyzed. To determine texture regions on images that are of a combined spectral-statistical nature, a hybrid texture identification method has been developed which makes it possible to take into account the local characteristics of the texture based on multifractal indicators characterizing the non-stationarity and impulsite of the data and the sign of the spectral texture. The stages of the developed hybrid texture identification method are: preprocessing; formation of the primary features vector; formation of the secondary features vector. The formation of the primary features vector is performed for the selected rectangular fragment of the image, in which the multifractal features and the spectral texture feature are calculated. To reduce the feature space at the stage of formation of the secondary identification vector, the principal component method was used. An experimental study of the developed hybrid texture identification method textures on model images of spectral, statistical, spectral-statistical textures has been carried out. The results of the study showed that the developed method made it possible to increase the probability of correct determination of the region of the combined spectral-statistical texture. The developed identification method was tested on images from Brodatz album of textures and images of wear zones of cutting tools, which are processed in intelligent systems of technical diagnostics. The probability of correctly identifying areas of spectral-statistical texture in the images of wear zones of cutting tools averaged 0.9, which is sufficient for the needs of practice.

An Efficient Text Detection Model for Street Signs

Text detection in natural scenes is a current research hotspot. The Efficient and Accurate Scene Text (EAST) detector model has fast detection speed and good performance but is ineffective in detecting long text regions owing to its small receptive field. In this study, we built upon the EAST model by improving the bounding box’s shrinking algorithm to make the model more accurate in predicting short edges of text regions; altering the loss function from balanced cross-entropy to Focal loss; improving the model’s learning ability on hard, positive examples; and adding a feature enhancement module (FEM) to increase the receptive field of the EAST model and enhance its detection ability for long text regions. The improved EAST model achieved better detection results on both the ICDAR2015 dataset and the Street Sign Text Detection (SSTD) dataset proposed in this paper. The precision and F1 scores of the model also demonstrated advantages over other models on the ICDAR2015 dataset. A comparison of the text detection effects between the improved EAST model and the EAST model showed that the proposed FEM was more effective in increasing the EAST detector’s receptive field, which indicates that it can improve the detection of long text regions.

Text Region Identification in Indian Street Scene Images Using Stroke Width Transform and Support Vector Machine

Text Region Identification in Indian Street Scene Images Using Stroke Width Transform and Support Vector Machine

Self-Supervised Monocular Depth Estimation of Untextured Indoor Rotated Scenes

Self-supervised deep learning methods have leveraged stereo images for training monocular depth estimation. Although these methods show strong results on outdoor datasets such as KITTI, they do not match performance of supervised methods on indoor environments with camera rotation. Indoor, rotated scenes are common for less constrained applications and pose problems for two reasons: abundance of low texture regions and increased complexity of depth cues for images under rotation. In an effort to extend self-supervised learning to more generalised environments we propose two additions. First, we propose a novel Filled Disparity Loss term that corrects for ambiguity of image reconstruction error loss in textureless regions. Specifically, we interpolate disparity in untextured regions, using the estimated disparity from surrounding textured areas, and use L1 loss to correct the original estimation. Our experiments show that depth estimation is substantially improved on low-texture scenes, without any loss on textured scenes, when compared to Monodepth by Godard et al. Secondly, we show that training with an application's representative rotations, in both pitch and roll, is sufficient to significantly improve performance over the entire range of expected rotation. We demonstrate that depth estimation is successfully generalised as performance is not lost when evaluated on test sets with no camera rotation. Together these developments enable a broader use of self-supervised learning of monocular depth estimation for complex environments.

Multi-scale selective image texture smoothing via intuitive single clicks

Existing texture filtering techniques acquiesce in the removal of all undesired background texture patterns simultaneously, and lead to the disappearance of small-scale foreground structures. Differently, this paper considers both multi-scale structures and specified texture patterns within an image as significant image contents, and presents a content-aware texture filtering approach via an easy-to-use interactive interface, named intuitive single click. The user’s effort on content-aware texture filtering is reduced to only locating a single click, where it is intuitively considered as one type of texture patterns. This simple interface is made possible by first generating three-scale intermediate filtered images (i.e., high, medium and low) and then performing breadth-first search (BFS) to identify one or more texture regions. Specifically, we utilize BFS to purify image edges in the medium-scale filtered image, on top of which a mask representing texture and non-texture regions is abstracted through the intuitive single-click texture locating and BFS. The mask is applied to effectively maintain multi-scale structures in the low-scale filtered image, and guides to fully smooth textures in the high-scale filtered image. Finally, a clean image is yielded by simply synthesizing the results of two mask operations in the low- and high-scale filtered images. We illustrate by extensive experiments that our approach advances the state-of-the-arts in terms of both the image cleanness and content-preserving ability. Our code is publicly available.

PDANet: Self-Supervised Monocular Depth Estimation Using Perceptual and Data Augmentation Consistency

In recent studies, self-supervised learning methods have been explored for monocular depth estimation. They minimize the reconstruction loss of images instead of depth information as a supervised signal. However, existing methods usually assume that the corresponding points in different views should have the same color, which leads to unreliable unsupervised signals and ultimately damages the reconstruction loss during the training. Meanwhile, in the low texture region, it is unable to predict the disparity value of pixels correctly because of the small number of extracted features. To solve the above issues, we propose a network—PDANet—that integrates perceptual consistency and data augmentation consistency, which are more reliable unsupervised signals, into a regular unsupervised depth estimation model. Specifically, we apply a reliable data augmentation mechanism to minimize the loss of the disparity map generated by the original image and the augmented image, respectively, which will enhance the robustness of the image in the prediction of color fluctuation. At the same time, we aggregate the features of different layers extracted by a pre-trained VGG16 network to explore the higher-level perceptual differences between the input image and the generated one. Ablation studies demonstrate the effectiveness of each components, and PDANet shows high-quality depth estimation results on the KITTI benchmark, which optimizes the state-of-the-art method from 0.114 to 0.084, measured by absolute relative error for depth estimation.

Depth Estimation Using a Self-Supervised Network Based on Cross-Layer Feature Fusion and the Quadtree Constraint

Depth estimation from a camera is an important task for 3D perception. Recently, without using the labeled ground truth of depth map, a self-supervised deep learning network can use relative pose to synthesize the target image from the reference image, and the photometric error between synthesized reference image and real one is used as self-supervisory signal. In this paper, we propose a novel self-supervised depth estimation network, which takes advantage of the quadtree constraint to optimize the depth estimation network. Based on the quadtree constraint, the photometric loss and depth loss of quadtree are proposed. In order to solve the problem that multiple depth values in repeated structures and uniform texture regions can cause relatively low photometric loss, we use quadtree-based photometric loss, which calculates the averaged photometric loss in quadtree blocks instead of the pixel-wise loss. For the problem of imbalanced depth distribution, we use quadtree depth loss, which constrains the depth inconsistency within quadtree blocks. The depth estimation network is composed of deep fusion module and cross-layer feature fusion module, which can better extract the feature information of RGB image and sparse keypoints depths, and makes full use of the detail information of the shallow feature map and the semantic information of the deep feature map to enrich the feature information extraction. Experimental results demonstrate that our method outperforms the state-of-the-art approaches of depth estimation.