(For USM Staff/Student Only)

EngLib USM > Ω School of Electrical & Electronic Engineering >

Development of example-based grayscale image colorization method

Development of example-based grayscale image colorization method / Chong Kean Sek
Pewarnaan adalah tugas memberikan warna kepada imej kelabu monokrom. Ia kebanyakannya diguna untuk mewarnakan fotografi monokrom lama atau bersejarah dan filem-filem klasik untuk meningkatkan daya tarikannya. Walau bagaimanapun, kaedah pewarnaan semasa masih memerlukan usaha intensif daripada pengguna untuk mencapai keputusan yang memuaskan. Oleh itu, satu kaedah pewarnaan berdasarkan contoh telah dicadangkan dengan pengguna hanya perlu untuk membekalkan imej warna rujukan yang semantiknya serupa dengan imej sasaran kelabu. Secara umumnya, kaedah yang dicadangkan terdiri daripada enam proses utama iaitu imej pra-pemprosesan, pengekstrakan ciri, segmentasi imej, pemadanan segmen, pemindahan warna & pewarnaan dan imej pasca pemprosesan. Terdapat beberapa ciri-ciri penting yang menjadikan kaedah ini berjaya. Yang pertama, analisis diskriminan linear yang dimasukkan ke dalam proses pemadanan segmen telah meningkatkan ketepatan pemadanan piksel antara imej kelabu dan warna dengan ketara. Selain itu, skim segmen mengundi menggabungkan dengan langkah warna penghalusan telah sebahagian besarnya membuangkan warna yang tidak sah. Ini menghasilkan imej berwarna berkualiti tinggi yang mempamerkan pertalian warna yang lebih kuat tanpa apa-apa artifak warna jelas. Keputusan telah menunjukkan bahawa kaedah yang dicadangkan adalah lebih baik daripada kaedah pewarnaan Welsh et al., Irony et al. dan Gupta et al. Ia mencatatkan purata isyarat puncak tertinggi kepada nisbah bunyi, 19.975 dengan peningkatan peratusan 1.96%, 20.39% dan 21.04% berbanding dengan kaedah perwarnaan Gupta et al., Irony et al. dan Welsh et al., masing-masing. _______________________________________________________________________________________________________ Colorization is a task of assigning colors to a monochrome gray image. It is mostly used in colorizing the old or historic monochrome photography and classic films to enhance its visual attractiveness. However, the current colorization methods still require intensive manual efforts from the user to achieve satisfactory results. Thus, an example-based colorization method has been proposed in which the user is only required to supply a reference color image which is semantically similar to the target gray image. Generally, the proposed method consists of six main processes which are image pre-processing, feature extraction, image segmentation, segment matching, color transfer & colorization and image post-processing. There are some important features which makes this method successful. First, linear discriminant analysis is incorporated into the segment matching process which improves the matching accuracy of pixels between gray and color image significantly. Besides, the segment voting scheme incorporates with color refinement step has largely eliminated the invalid color assignments. These results in a high quality colorized image which exhibits much stronger spatial color coherency without any apparent color artifacts. Results have shown that the proposed method is better than the colorization method of Welsh et al., Irony et al. and Gupta et al. It recorded a highest average peak signal to noise ratio of 19.975 with improvement percentage of 1.96%, 20.39%, and 21.04% over the colorization method of Gupta et al., Irony et al. and Welsh et al., respectively
Contributor(s):
Chong Kean Sek - Author
Primary Item Type:
Final Year Project
Identifiers:
Accession Number : 875006081
Language:
English
Subject Keywords:
Colorization; assigning colors; monochrome gray image
First presented to the public:
6/1/2016
Original Publication Date:
6/14/2018
Previously Published By:
Universiti Sains Malaysia
Place Of Publication:
School of Electrical & Electronic Engineering
Citation:
Extents:
Number of Pages - 111
License Grantor / Date Granted:
  / ( View License )
Date Deposited
2018-06-14 12:04:33.48
Date Last Updated
2019-01-07 11:24:32.9118
Submitter:
Mohd Jasnizam Mohd Salleh

All Versions

Thumbnail Name Version Created Date
Development of example-based grayscale image colorization method1 2018-06-14 12:04:33.48