Text Region Conditional Generative Adversarial Network for Text Concealment in the Wild

The paper deals with a suitably designed system that is being used to separate textual regions from graphics regions and locate textual data from textured background. We presented a method based on edge detection to automatically locate text in some noise infected grayscale newspaper images with microfilm format. The algorithm first finds the appropriate edges of textual region using Canny edge detector, and then by edge merging it makes use of edge features to do block segmentation and classification, afterwards feature aided connected component analysis was used to group homogeneous textual regions together within the scope of its bounding box. We can obtain an efficient block segmentation with reduced memory size by introducing the TLC. The proposed method has been used to locate text in a group of newspaper images with multiple page layout. Initial results are encouraging, we would expand the experiment data to over 300 microfilm images with different layout structures, promising result is anticipated with corresponding modification on the prototype of former algorithm to make it more robust and suitable to different cases.

In this paper, we propose a new method for natural image text detection under a weakly supervised data set. Currently, most of the text detection models are based on bounding box label training data. However, the cost of the bounding box label training data is very high. In order to solve this problem, we propose an attention mechanism that can be trained on image-level labels data and roughly identifies text regions via an automatically learned attentional map based on a convolutional neural network. There are three main steps: firstly, a VGG model is trained using image-level labels data to score the likelihood that a text region exists in the picture; secondly, the region of interest is extracted by means of the attention mechanism and the extracted region is evaluated using the network trained in the first step to getting the text region and finally, the text line is extracted in the text region using the MSER algorithm. Trained with the weakly supervised data which is only with image-level labels, our model can generate bounding boxes for the text line in the image. The results of our model are very close to those of the models using bounding box label training data on the text detection benchmark sets of MSRA-TD500, ICDAR2013, and ICDAR2015.

Adaptive background updating is an important step in motion segmentations of video sequences. However, the irregular distributions of background pixel values make the background modeling complicated. In this work, a method for background pixel classification based on the mean shift algorithm is proposed, which can classify the background pixels as single mode or multiple mode pixels so that different updating methods can be applied, for example, the single-mode pixel values can be updated with a simple and fast method such as IIR filter, while the multi-mode pixel values are modeled by a more complex updating algorithm such as a mixture of K Gaussian distributions or the non-parametric kernel estimation, which is robust to small motions (noisy motions, repetitive motions). Since in most scenes, the number of static background pixels (single-mode pixels) is far more than the number of dynamic background pixels (multi-mode pixels), thus the presented method can help improve the speed of background reconstruction without reducing its precision.

Object tracking is one of the most important topics in computer vision. While the state-of-the-art tracking algorithms achieved great success, there are still some challenging problems to be solved. Firstly, it remains a tough task to develop a tracking algorithm with both accuracy and efficiency. Secondly, the ground truth is often given by a rectangular bounding box, which contains not only the target, but also the background pixels. The background pixels in the initial bounding box will mislead the appearance model of the target. Thirdly, most features for representing the target only use the gray scale information, and are not robust to pedestrians. The third problem is especially serious when the targets are pedestrians, which have colorful cloths on them, and change their pose and shape while walking. In this paper, we propose a novel saliency-based H-S histogram algorithm. Saliency detection can efficiently delete most of the background pixels, which makes the appearance model more accurate. H-S histogram is a statistic-based feature, which is more robust to shape deformation, and the color information is considered in the H-S channels. Experiments on Caltech pedestrian database show the proposed method can handle some hard cases, and achieves a higher success rate.

Text detection in natural and complex images play an important role in analysis of image. Text information appears everywhere like product labels, documents, scene images etc. It will be very difficult for visually impaired person to find text region in an image. This problem should be addressed because it cannot assume that the captured image contains only text. This paper presents camera-based text reading framework which helps visually impaired people to read texts in natural scenes, product labels etc. Here, task is divided into two steps i.e. text detection and text recognition. Text detection involves text localization phase. Firstly, we binarize the image. Geometric and stroke width filtering removes non text regions an image. Segmented text region in a cluttered scene is binarized and recognized by optical character recognition. Recognized texts are output to the visually impaired people in speech. Experimental result shows that proposed method offers better performance in text recognition.

The advantage of spatial domain image steganography techniques is their capacity to embed high payloads of data by directly modifying image pixels. While these techniques have a high-embedding capacity, they often create visual and statistical distortion in smoother regions. Most existing edge steganography techniques divide an image into blocks and insert data by processing the blocks in a linear order, but these method also has multiple drawbacks. First, if the selected block has an insufficient number of edge pixels, it may result in multiple blocks being processed. Second, at high embedding rates, the method creates severe distortion as multiple message bits are hidden in edge pixels and surrounding non-edge pixels without analyzing the statistical dependencies and correlation of pixels, compromising data security. The aim of the proposed method is to construct a Block-wise Edge Adaptive Steganography Scheme (BEASS) using textured regions, particularly edges and surrounding pixels. This scheme dynamically chooses the region to embed messages using a local complexity measure of Standard Deviation. It offers high payload, minimal distortion embedding by hiding three message bits into edge pixels using the minimal Mean Square Error to determine the embedding capacity of neighboring non-edge pixels within the block to preserve the statistical dependencies. The practical merit of this approach was validated and compared with existing algorithms, and experimental results find that the proposed method surpasses IQM tests, achieves a high PSNR of61 $\sim $ 65, proves to be robust against kurtosis and skewness distortion, resists histogram attack, RS steganalysis and high dimensional ensemble classifier at 80% block modifications.

Textual data plays an important role in a number of applications such as image database indexing, document understanding, and image-based web searching. The target of automatic real-life text extracting in document images without character recognition module is to identify image regions that contain only text. These textual regions can then be either input of optical character recognition application or highlighted for user focusing. In this paper we propose a method which consists of three stages-preprocessing which improves contrast of grayscale image, multi-level thresholding for separating textual region from non-textual object such as graphics, pictures, and complex background, and heuristic filter, recursive filter for text localizing in textual region. In many of these applications, it is not necessary to identify all the text regions, therefor we emphasize on identifying important text region with relatively large size and high contrast. Experimental results on real-life dataset images demonstrate that the proposed method is effective in identifying textual region with various illuminations, size and font from various types of background.

A deep texture adaptive denoising method is proposed to achieve high perceptual image quality. Textual information is learned through a designed loss function utilizing a pre-generated texture map to distinguish textual areas from flat areas. In the training process, the proposed network internally finds texture and flat regions and differs in denoising strength in the two regions. Unlike existing DNN-based denoising methods, the proposed method retains high-frequency textual information while removing residual noise in flat regions as much as possible. The gradient distribution of the image before and after the denoising was compared. The proposed method outperformed the existing methods with higher PSNR and SSIM scores in visual quality. In addition, the strength of removing textual noise was controllable with a single parameter. Thus, the proposed method is practically feasible as a denoising apparatus.

Now days reading words from an unconstrained and noisy image is not easy. Text localization and recognition in an image is a research area which takes efforts to develop a computer system with an ability to automatically read the text from images. The Optical Character Recognition (OCR) tool gives good results obtained to read the text from an image. The objective of this study is to propose a new method for text localization and recognition in natural scene images with complex background. In this paper, a hybrid methodology is suggested which extracts text from natural scene image with chaotic backgrounds. The proposed approach involves four stages. First, superimposed text regions in an image are extracted based on character descriptors features like Area, Bounding box, Perimeter, Euler number, Horizontal crossings. In the second step, superimposed text regions are tested for text content or nontext using character descriptors and SVM classifier. In the third step, detection of multiple lines in localized text regions is done and line segmentation is performed using horizontal profiles. In the final step, using vertical profiles each character of the segmented line is extracted. The workout has been done using images drawn from ICDAR 2013 and SVT 2010 datasets. The results demonstrate the effectiveness of the proposed method, which can be used as an efficient method for text localization and recognition in natural scene images.

Corner detection is a fundamental step for many computer vision applications to detect the salient image features. Recently, FAST corner detector has been proposed to detect the high repeatable corners with efficient computational time. However, FAST is very sensitive to noise and detects too many unnecessary corners on the noise or texture region. In this paper, we propose a robust corner detector improved from FAST in terms of the localization accuracy and the computational time. First, we construct a gradient map using the Haar-wavelet response by integral image for efficiency. Second, we define a corner candidate region which has large gradient magnitude enough to be corner. Finally, we detect the corner on the corner candidate region by FAST. Experimental results show the proposed method improves localization accuracy measured by the repeatability than standard FAST and the-state-of-art methods. Moreover, the proposed method shows the best computation efficiency. Especially, the proposed method detects the corners more accurately in the image containing many texture regions and corrupted by the Gaussian noise or the impulse noise.

Bin loss for hard exudates segmentation in fundus images

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.

SP21 Needle architecture: vulnerability of different nerves to intrafascicular injection with different needle bevels

In this paper we present a well designed method that makes use of edge information to extract textual blocks from gray scale document images. It aims at detecting textual regions on heavy noise infected newspaper images and separate them from graphical regions. The algorithm traces the feature points in different entities and then groups those edge points of textual regions. From using the technology of line approximation and layout categorization, it can successfully retrieve directional placed text blocks. Finally feature based connected component merging was introduced to gather homogeneous textual regions together within the scope of its bounding rectangles. We can obtain correct page decomposition with efficient computation and reduced memory size by handling line segments instead of small pixels. The proposed method has been tested on a large group of newspaper images with multiple page layouts, promising results approved the effectiveness of our method.

We propose a new method to extract foreground edges in a video streams taken from a stationary camera. Our background model is based on the fact that a background pixel's gradient components follow Gaussian mixture model(GMM). GMM is performed on the initial group of video frames to obtain the initial pixel gradient component distribution information at each pixel. Then each of the current Canny edge pixels is classified into foreground or background pixel based on its gradient components' weighted square sum of distances from their respective mean values. If the difference is larger than a threshold, it is then classified as a foreground pixel, otherwise a background pixel in which case the GMM information is accordingly updated. If the ratio of the number of foreground pixels over the total number of Canny edge pixel is large than a certain threshold, a new GMM background modeling is trigger. The algorithm is implemented in Visual C++ and tested on a laptop powered by an Intel Pentium 3.0GHz. The experiment shows the algorithm is highly selective in extracting valid foreground edge pixels and it's speed is 43 ms/frame for a video stream of 640×480 and shows that the method is applicable for real-time processing.