Full Record

New Search | Similar Records

Author
Title Text Detection using Coarse detection and SVM Classification
URL
Publication Date
Degree Level masters
University/Publisher Universiteit Utrecht
Abstract In this paper a new approach to detect text in natural images is described using different detection methods. The end result will be that the text will be segmented from the image and can be used for different purposes. The approach is split up in two parts, a coarse detection step to extract patches from the image and a fine detection step that uses feature descriptors and a support vector machine in order to increase the precision of the coarse detection step. The methods used for the coarse detection are global threshold, mean threshold, Gaussian threshold, local binary pattern, maximum gradient difference filter and maximum difference filter. These methods are compared and the best results are used in combination with the fine detection. The feature descriptors used in the fine detection are Histogram of Oriented Gradients, Co-occurrence histogram of orientated gradients and local binary patterns. In order to increase the quality of the coarse detection a projection step is used. The approach performs on precision level worse than the current state-of-the-art methods, but has a better recall rate than most methods.
Subjects/Keywords Text detection, Histogram of Orientated Gradients, Coarse-to-fine schema, Support Vector Machines, thresholds
Contributors Veltkamp, R; Tan, R; Fu, Z
Language en
Rights info:eu-repo/semantics/OpenAccess
Country of Publication nl
Format image/pdf
Record ID oai:dspace.library.uu.nl:1874/296615
Repository utrecht
Date Indexed 2016-09-27

Sample Search Hits | Sample Images

…for the next step. The last step of the method is to define the text areas. This part consists of two phases: a coarse phase and a fine phase, both split up in a horizontally projection and vertically projection step. After these two steps the regions…

…different fonts, font sizes, background complexity and contrast levels. Also a coarse-to-fine detection schema is used in order to detect text. The major downside for the problem this paper tries to solve, is that it is focused on text detection in videos…

…namely coarse-to-fine detection schema [4] can be used in order to find decent results, very cheap feature descriptors [14] can be used in order to do the fine detection and the different aspects of text which are described in section…

…characters can be used in order to detect text in images as has been shown in [2]. 9 Chapter 3 Our approach In order to solve the problem described in the introduction a coarse-to-fine pipeline is introduced. The reason for using a coarse-to

fine pipeline is that by using a coarse step to start with a selection of interesting pixels can be found and easily segmentated in order to create feature descriptors for the second part. The precision of the coarse detection will be low, hence why a…

…the fine detection that improves the precision of the results patches of the image are extracted. These patches have been marked by the coarse detection as text pixels. For each of these patches a projection step was used in order to split up the…

fine detection schema that can be used as a base for future work that is cheap. Also it is shown that the CoHoG feature descriptor can be used for different purposes. A strength of the coarse-to-fine detection approach is that the problem of detecting…

…can be experimented with as well. This feature descriptor has been used without any coarse detection step with decent results on precision level. The coarse-to-fine detection has shown that it works pretty well, however an extra elimination step could…

.