Hong Zhao

Bio: Hong Zhao is an academic researcher from College of Information Technology. The author has contributed to research in topics: Image segmentation & Conditional random field. The author has an hindex of 2, co-authored 4 publications receiving 15 citations.

Research on the Text Detection and Extraction from Complex Images

Abstract: The text appears in the images is important for fully understanding the images. The number of digital images and digital videos has increased tremendously. Although there are many methods have been proposed over the past years for the text extraction from natural scene images, the text detection and extraction from born-digital images are still a challenge. In this paper, we describe existing methods key ideas and try to summarize their advantages and disadvantages. We try to find a new way which can Comprehensive utilize existing methods to detect and extract text from born-digital image.

Research on Born-Digital Image Text Extraction Based on Conditional Random Field

Abstract: Born-digital images are generated directly with the computer, the text in the images is important for fully understanding the images. Although there are many methods having been proposed over the past years for text extraction from natural scene images, text detection and extraction from born-digital images are still a challenge. This paper proposed an algorithm of text extraction from born-digital images based on conditional random field (CRF). CRF model not only considers unary component properties and binary contextual component relationships, but also learn parameter s with supervised. This paper combines features and relationships within the CRF framework and the experiment results show that this algorithm can extract text effectively from the born-digital images.

Optimization of Air Pollutant Monitoring Stations Based on Genetic Algorithm

Abstract: Reasonable sitting an air quality monitoring stations (AQMS) network is an important task for environmental protection department. However, involving many factors, optimizing air quality monitoring sites has been proven to be in the class of nondeterministic polynomial (NP)-hard problem. The powerful search capability of the genetic algorithm (GA) is a key factor in improving the performance of selecting optimal monitoring sites. A mathematical programming model is proposed, which the GA is used in order to optimize AQMS. The environmental, social, and economic objectives are considered in the optimization model. Modelling results suggest that the proposed approach outperforms the method of random site of AQMS.

A Comparative Study on CRFs for Fore- and Background Classification

Abstract: Text contained in images and video frames provide an important clue for content based images and video indexing. However, in natural scene, text elements are corrupted by many types of noise, such as streaks, highlights, or cracks. These effects make the clean and automatic segmentation very difficult and can reduce the accuracy of further analysis such as optical character recognition. This paper uses different optimization methods for text segmentation from images with complex background based on conditional random field. Experimental results demonstrate the performance of the different methods. It enables the segmentation of text in complex situations more precisely comparing to the binarization of the image.

A comprehensive of transforms, Gabor filter and k-means clustering for text detection in images and video

Abstract: The present paper presents one of the efficient approaches toward multilingual text detection for video indexing. In this paper, we propose a method for detecting textlocated in varying and complex background in images/video. The present approach comprises four stages: In the first stage, combination of wavelet transform and Gabor filter is applied. By applying single level 2D wavelet decomposition with Gabor Filter, the intrinsic features comprising sharpen edges and texture features of an input image are obtained. In the second stage, the resultant Gabor image is classified using k-means clustering algorithm. In the third stage, morphological operations are performed on clustered pixels. Then a concept of linked list approach is used to build a true textline sequence of connected components. In the final stage, wavelet entropy of an input image is measured by signifying the complexity of unsteady signals corresponding to the position of textline sequence of connected components in leading to determine the true text region of an input image. The performance of the approach is exhibited by presenting promising experimental results for 101 video images, standard ICDAR 2003 Scene Trial Test dataset, ICDAR 2013 dataset and on our own collected South Indian Language dataset.

Text detection and localization in scene images: a broad review

Abstract: Nowadays, text detection and localization have gained much popularity in the field of text analysis systems as they pave the way for the number of real-time based applications like mobile transliteration technologies, assistive methods for visually impaired persons, etc. Text detection and localization techniques are used to find the position of text area in the image.This paper intends to present a broad review in this field as five-fold: (1) comparison of document images with scene images and applications of natural scene images, (2) significant and up-to-date traditional machine learning and deep learning-based approaches for the text detection and localization for different languages, (3) various publicly available benchmarked datasets, (4) comparative analysis for other benchmarked datasets and, (5) related challenges and future scope on the field. The paper summarises some of the potential ways in this field, which can serve as a useful reference for the researchers for future exploration of the area.

Analysis and optimal design of air quality monitoring networks using a variational approach

Abstract: Air quality networks need revision and optimisation as instruments and network requirements, both scientific and societal, evolve over time. Assessing and optimising the information content of a monitoring network is a non-trivial problem. Here, we introduce a methodology formulated in a variational framework using an air quality model to simulate the dispersion of carbon monoxide (CO) as a passive tracer at the city scale. We address the specific case of adding or removing stations, and the more general situation of optimally distributing a given number of stations in a domain taking into account transport patterns and spatial factors such as population density and emission patterns. We consider three quality indicators: precision gain, information gain and degrees of freedom for a signal. These metrics are all functions of the singular values of the sensitivity matrix that links emissions and observations in the variational framework. We illustrate the application of the methodology in the case of Santiago (33.5°S, 70.5°W, 500 m a.s.l.), a city of ca. 7 million inhabitants with significant pollution levels. We deem information gain as the best of the above indicators for this case. We then quantify the actual evolution of Santiago’s network and compare it with the optimal configuration suggested by our methodology and with results previously obtained using a statistical approach. The application is restricted to diurnal and summer conditions, for which the dispersion model shows a good agreement with observations. The current method offers advantages in that it allows extending a network to include new sites, and it explicitly considers the effects of dispersion patterns, and desired weighting functions such as emission fluxes and population density. We find that Santiago’s air quality has improved two-fold since 1988, regarding CO under diurnal summer conditions. Still, according to our results, the current configuration could be improved by integrating more suburban stations in the southwest of the basin. Keywords: optimal network design, air pollution, variational framework (Published: 19 October 2015) Citation: Tellus B 2015, 67, 25385, http://dx.doi.org/10.3402/tellusb.v67.25385

Survey on Text Detection, Segmentation and Recognition from a Natural Scene Images

Abstract: Detecting text from an image is an important prerequisite for the content based image analysis process. To understand the contents of an image or the valuable information, there is need of analyzing the text appears in it. Various methods have been proposed over past years for text detection and extraction from different types of images, like scene image, born digital image and document image. In this paper, we describe the existing methods of text detection, text segmentation and character recognition from natural scene images with their features, advantages and disadvantages. General Terms Pattern Recognition

Layout and Text Extraction from Document Images using Neural Networks

Abstract: This paper attempts to provide a new perspective for efficient text extraction techniques by including the layout information of the document images along with the text. The study proposes a novel approach of segmenting the image into smaller images based on its meta-data knowledge and then applying functions for recognition of text from the smaller images. Due to a lack of layout information of the text, poor results are generated during text searching by office automation tools. With this restriction usage of the text for various environments becomes limited and usage of extracted text may not be done effectively. The study proposes a technique to understand the structural and functional layout of the document image and using this knowledge to develop a better model. To verify the point of view, additional intelligence is attached to the data, making it capable to be used in varied environments. With this added quality, the new proposed system can extract and identify text or group of text into different entities within the document which the previous systems could not achieve. The proposal can be beneficial particularly for the development of various document processing tools.