TEXT LOCALIZATION IN IMAGES USING REVERSE THRESHOLDS ALGORITHM

2015 ◽  
Vol 75 (2) ◽  
Author(s):  
Lih-Fong Wong ◽  
Mohd Yazid Idris ◽  
Abdul Hanan Abdullah
Keyword(s):  
Similarity Index ◽  
Background Color ◽  
Text Localization ◽  
Color Similarity ◽  
Text Features ◽  
Rough Calculation ◽  
Novel Algorithm

High color similarity between text pixels and background pixels is the major problem that causes failure during text localization. In this paper, a novel algorithm, Reverse Thresholds (RT) algorithm is proposed to localize text from the images with various text-background color similarities. First, a rough calculation is proposed to determine the similarity index for every text region. Then, by applying reverse operation, the best thresholds for each text region are calculated by its similarity index. To remove other uncertainties, self-generated images with the same text features but different similarity index are used as experiment dataset. Experiment result shows that RT algorithm has higher localizing strength which is able to localize text in a wider range of similarity index.

scholarly journals Assessment of Camouflage Effectiveness Based on Perceived Color Difference and Gradient Magnitude

Sensors ◽  
2020 ◽  
Vol 20 (17) ◽  
pp. 4672
Author(s):  
Xueqiong Bai ◽  
Ningfang Liao ◽  
Wenmin Wu
Keyword(s):  
Real World ◽  
Information Entropy ◽  
Human Visual System ◽  
Similarity Index ◽  
Color Difference ◽  
Entropy Theory ◽  
Analysis Process ◽  
Gradient Magnitude ◽  
Color Similarity ◽  
Perceived Color

We propose a new model to assess the effectiveness of camouflage in terms of perceived color difference and gradient magnitude. The “image color similarity index” (ICSI) and gradient magnitude similarity deviation (GMSD) were employed to analyze color and texture differences, respectively, between background and camouflage images. Information entropy theory was used to calculate weights for each metric, yielding an overall camouflage effectiveness metric. During the analysis process, both spatial and color perceptions of the human visual system (HVS) were considered, to mimic real-world observations. Subjective tests were used to compare our proposed method with previous methods, and our results confirmed the validity of assessing camouflage effectiveness based on perceived color difference and gradient magnitude.

Idea plagiarism detection with recurrent neural networks and vector space model

2021 ◽  
Vol ahead-of-print (ahead-of-print) ◽  
Author(s):  
Azra Nazir ◽  
Roohie Naaz Mir ◽  
Shaima Qureshi
Keyword(s):  
Similarity Index ◽  
Similarity Score ◽  
Academic Community ◽  
Plagiarism Detection ◽  
Natural Languages ◽  
Content Type ◽  
Proposed Model ◽  
Text Features ◽  
Single Idea

PurposeNatural languages have a fundamental quality of suppleness that makes it possible to present a single idea in plenty of different ways. This feature is often exploited in the academic world, leading to the theft of work referred to as plagiarism. Many approaches have been put forward to detect such cases based on various text features and grammatical structures of languages. However, there is a huge scope of improvement for detecting intelligent plagiarism.Design/methodology/approachTo realize this, the paper introduces a hybrid model to detect intelligent plagiarism by breaking the entire process into three stages: (1) clustering, (2) vector formulation in each cluster based on semantic roles, normalization and similarity index calculation and (3) Summary generation using encoder-decoder. An effective weighing scheme has been introduced to select terms used to build vectors based on K-means, which is calculated on the synonym set for the said term. If the value calculated in the last stage lies above a predefined threshold, only then the next semantic argument is analyzed. When the similarity score for two documents is beyond the threshold, a short summary for plagiarized documents is created.FindingsExperimental results show that this method is able to detect connotation and concealment used in idea plagiarism besides detecting literal plagiarism.Originality/valueThe proposed model can help academics stay updated by providing summaries of relevant articles. It would eliminate the practice of plagiarism infesting the academic community at an unprecedented pace. The model will also accelerate the process of reviewing academic documents, aiding in the speedy publishing of research articles.

Bleeding Detection in Wireless Capsule Endoscopy Using Color Similarity Coefficient

2012 ◽  
Vol 195-196 ◽  
pp. 307-312 ◽  
Author(s):  
Guo Bing Pan ◽  
Fang Xu ◽  
Jiao Liao Chen
Keyword(s):  
Capsule Endoscopy ◽  
Color Space ◽  
Wireless Capsule Endoscopy ◽  
The Novel ◽  
Similarity Coefficients ◽  
Bleeding Detection ◽  
Color Similarity ◽  
Color Vector ◽  
Wireless Capsule ◽  
Novel Algorithm

Wireless Capsule Endoscopy (WCE) generates a large number of images in one examination of a patient. It is very laborious and time-consuming to detect the WCE video, and limits the wider application of WCE. It is urgent and necessary to develop an automatic and intelligent computer aided bleeding detection technique. This paper proposes the color vector similarity coefficients to measure the color similarity, and based on which, a novel algorithm is implemented to recognize the bleeding in WCE images. The novel algorithm is implemented in RGB color space, and is featured with simple computation and practicability. The experiments show the sensitivity and specificity of this algorithm are 90% and 97% respectively.

The Effect of Symbol Background Color on the Speed of Locating Targets by Adults Without Disabilities: Implications for Augmentative and Alternative Communication Display Design

2019 ◽  
Vol 4 (6) ◽  
pp. 1482-1488
Author(s):  
Jennifer J. Thistle
Keyword(s):  
Augmentative And Alternative Communication ◽  
Reaction Times ◽  
The Other ◽  
Background Color ◽  
Alternative Communication ◽  
Display Design ◽  
Line Drawings ◽  
Research With Children ◽  
Study Participants ◽  
Perceptual Factors

Purpose Previous research with children with and without disabilities has demonstrated that visual–perceptual factors can influence the speech of locating a target on an array. Adults without disabilities often facilitate the learning and use of a child's augmentative and alternative communication system. The current research examined how the presence of symbol background color influenced the speed with which adults without disabilities located target line drawings in 2 studies. Method Both studies used a between-subjects design. In the 1st study, 30 adults (ages 18–29 years) located targets in a 16-symbol array. In the 2nd study, 30 adults (ages 18–34 years) located targets in a 60-symbol array. There were 3 conditions in each study: symbol background color, symbol background white with a black border, and symbol background white with a color border. Results In the 1st study, reaction times across groups were not significantly different. In the 2nd study, participants in the symbol background color condition were significantly faster than participants in the other conditions, and participants in the symbol background white with black border were significantly slower than participants in the other conditions. Conclusion Communication partners may benefit from the presence of background color, especially when supporting children using displays with many symbols.

A Test of Three Sadism Measures

2020 ◽  
Vol 41 (4) ◽  
pp. 219-227 ◽  
Author(s):  
Bojana M. Dinić ◽  
Tara Bulut Allred ◽  
Boban Petrović ◽  
Anja Wertag
Keyword(s):  
Item Response ◽  
Social Groups ◽  
Similarity Index ◽  
Dark Triad ◽  
Hierarchical Regression ◽  
Fit Indices ◽  
Profile Similarity ◽  
Positive Attitudes ◽  
Confirmatory Factor ◽  
Positive Correlations

Abstract. The aim of this study was to evaluate psychometric properties of three sadism scales: Short Sadistic Impulse Scale (SSIS), Varieties of Sadistic Tendencies (VAST, which measures direct and vicarious sadism), and Assessment of Sadistic Personality (ASP). Sample included 443 participants (50.1% men) from the general population. Reliability based on internal consistency of all scales was good, and results of Confirmatory Factor Analysis (CFA) showed that all three scales had acceptable fit indices for the proposed structure. Results of Item Response Theory (IRT) analysis showed that all three scales had higher measurement precision (information) in above-average scores. Validity of the scales was supported through moderate to high positive correlations with the Dark Triad traits, especially psychopathy, as well as positive correlations with aggressiveness and negative with Honesty-Humility. Moreover, results of hierarchical regression analysis showed that all three measures of direct, but not vicarious sadism, contributed significantly above and beyond other Dark Triad traits to the prediction of increased positive attitudes toward dangerous social groups. The profile similarity index showed that the SSIS and the ASP were highly overlapping, while vicarious sadism seems distinct from other sadism scales.

Red, White, & Blue: Does Background Color Affect Solving Anagrams?

2011 ◽  
Author(s):  
Kenneth M. Steele ◽  
Christopher Thorstenson ◽  
Kristen Sugg ◽  
Erin Gurgainous ◽  
Andrea Steche ◽  
...  
Keyword(s):  
Background Color
Sign in / Sign up

Export Citation Format

Share Document