The Extraction of Characters on Dated Color Postcards

碩士 === 淡江大學 === 資訊工程學系 === 91 === In this study, we propose a scheme to extract characters from color postcards published around 1930 during the time Japan occupied Taiwan as one of its colonies. These postcards, reproduced in the book “台灣紀行” [1], possibly the earliest postcards in Taiwan...

Full description

Bibliographic Details
Main Authors: Mei-Fen Chen, 陳梅芬
Other Authors: Shwu-Huey Yen
Format: Others
Language:zh-TW
Published: 2003
Online Access:http://ndltd.ncl.edu.tw/handle/02708156348565075519
Description
Summary:碩士 === 淡江大學 === 資訊工程學系 === 91 === In this study, we propose a scheme to extract characters from color postcards published around 1930 during the time Japan occupied Taiwan as one of its colonies. These postcards, reproduced in the book “台灣紀行” [1], possibly the earliest postcards in Taiwan, contained photographs of beautiful mountains, rivers, farms, country view, famous buildings…etc of earlier Taiwan. As they are precious but fragile historical documents, it is very meaningful if these postcards can be kept in a digital library via key words searching engine, with helps of characters extraction, then these documents not only are available to researchers as well as to the general public. The illustrations printed on postcards usually are located either on the upper or lower positions of the postcards with various backgrounds. Characters may appear in black, white, or red with underneath backgrounds of blue sky and/or white clouds, dark mountains, brownish rocks, trees with leaves and/or branches, rocky or muddy country roads, ponds, etc. These characters also suffered problem of uneven illuminations. It will be a very difficult task if we want to find a global criterion for locating these characters for all postcards. To solve this problem, we use a morphological operation, reconstruction by erosion or dilation, followed by image subtraction to remove backgrounds connecting to borders of the postcards. As a result, characters become the most obvious objects in the image for most of irrelevant backgrounds being removed. Next, by vertical and horizontal projections and connected components analysis to remove noises, we can find the exact locations of the characters. Finally, back to the corresponding positions of the characters on color postcards, color contrast enhancement and sharpening are done as the post processing. The proposed algorithm has been executed on a set of color images of postcards and proved its efficacy.