ارسل ملاحظاتك

ارسل ملاحظاتك لنا







الرئيسية > Framework for Evaluating Scene...>الوصف

Framework for Evaluating Scene-Text Extraction Algorithms

المؤلف الرئيسي: Al Mutiry, Riyadh (Author)
مؤلفين آخرين: Zutshi, Samar (Advisor) , Wilson, Campbell (Advisor)
التاريخ الميلادي: 2011
موقع: ملبورن
الصفحات: 1 - 75
رقم MD: 752649
نوع المحتوى: رسائل جامعية
اللغة: الإنجليزية
الدرجة العلمية: رسالة ماجستير
الجامعة: Monash University
الكلية: Faculty of Information Technology
الدولة: أستراليا
قواعد المعلومات: +Dissertations
مواضيع:
الصور الرقمية | معالجة الصور | استخراج النصوص | الخوارزميات
رابط المحتوى:
صفحة العنوان     
المستخلص     
قائمة المحتويات     
24 صفحة الأولى     

الناشر لهذه المادة لم يسمح بإتاحتها.
صورة الغلاف QR قانون
حفظ في:
المستخلص: This project forms part of a larger research initiative regarding text extraction from digital images. In recent years, significant advances have been made in the automatic identification of candidate regions within images that could contain text. This represents an exciting intersection between image processing and artificial intelligence, with wide application to a variety of domains.This project involves firstly a comprehensive survey of relevant techniques and tools and secondly the implementation of an algorithm-independent framework for extracting text from images. It also, includes issues found in the literature regarding the evaluation of different algorithms. A new proposed challenge classification model which attaches semantics to both algorithms and images is adopted by the framework and proved to overcome major issues related to algorithm evaluation validity. These semantics can define the algorithm objectives and text characteristics found in the images database. The framework can be used to evaluate one or more text extraction techniques in terms of their efficacy for a particular application domain. This framework can serve as a testing harness for other applications and researchers. Some deficiencies are found after examining contour tracing algorithm used by the implemented text extraction algorithm. Some of these deficiencies were resolved to enhance the contour tracer performance in detecting object’s contour. However, some deficiencies remain challenging and future research is needed to overcome these issues.

عناصر مشابهة


© 2024 المنظومة. جميع الحقوق محفوظة.