(For USM Staff/Student Only)

EngLib USM > Ω School of Electrical & Electronic Engineering >

Evaluation of super-resolution methods for grayscale image/Cheah Tatt Wei

Evaluation of super-resolution methods for grayscale image_Cheah Tatt Wei_E3_2014_NI
Resolusi imej adalah jumlah maklumat yang terkandung dalam imej. Semakin tinggi resolusi, semakin banyak maklumat yang terkandung dalam imej. Oleh itu, imej berresolusi tinggi sering dikehendaki untuk pelbagai aplikasi pengimejan kerana kandungan maklumat yang tinggi dapat diperolehi daripadanya. Satu kaedah untuk meningkat resolusi imej adalah memadatkan sensor imej. Ini dapat dilakukan dengan mengurangkan saiz sensor imej tetapi mengurangkan kandungan cahaya kepada sensor yang akan mengurangkan imej quality. Kaedah ini juga berkos tinggi. Oleh kerana masalah ini timbul, kaedah super resolusi cuba untuk menyelesaikan had perkakasan dengan menggunakan pendekatan perisian. Interpolasi tidak seragam adalah salah satu kaedah penyelesaian super digunakan untuk projek ini. Selain daripada itu, kaedah interpolasi imej tunggal dan kaedah penambahan langsung dijalankan ke atas semua imej ujian juga. Perbandingan semua kaedah akan memberi gambaran tentang kepada kegunaan praktikal kaedah super resolusi. Dua belas imej telah dipilih sebagai imej ujian supaya kaedah imej kedua-dua satu dan berbilang itu dilaksanakan dan keputusan yang diperolehi dapat dianalisis dan dibandingkan. Keputusan menunjukkan nilai purata SSIM untuk kaedah jiran terdekat, bilinear, bicubic, penambahan langsung menggunakan min, penambahan langsung menggunakan median dan interpolasi tidak seragam untuk dua belas imej ujian (peratusan bunyi, , pensampelan faktor, and bilangan imej, ) adalah 0.921, 0.957, 0.968, 0.907, 0.905 dan 0.976 masing-masing. Ini menunjukkan kaedah interpolasi tidak seragam menghasilkan imej yang paling tinggi kualiti berbanding kaedah konvensional lain. ______________________________________________________________________________________ Image resolution is the detail an image holds. The higher the resolution, the more information that image holds. Hence, high resolution images are often desired for many imaging application for the vast amount of information obtained from it. One of the solutions to increase current spatial resolution is to increase sensor density. This can be done by reducing the size of sensor but with a drawback that is amount of light falling onto the sensors will reduce, hence causing shot noise which degrades the image quality severely. Increase sensor density also comes with high price. Due to the problem arises, super-resolution method attempts to solve hardware limitation by using software approach. This saves cost and preserves the usage of current low resolution imaging system. Non-uniform interpolation is one of the super resolution methods used for this project. Other than that, single image interpolation method and direct addition method are performed on all test images as well. Comparison of all methods will provide an insight to the practical uses of the super-resolution method. Twelve images were chosen as test image so that both single and multiple image methods are performed and the results obtained are analyzed and compared. Result shows that the average SSIM values for nearest neighbor, bilinear, bicubic, direct addition using mean, direct addition using median and non-uniform interpolation method for twelve test image (noise percentage, , sampling factor, and number of images, ) are 0.921, 0.957, 0.968, 0.907, 0.905 dan 0.976 respectively. The result shows the non-uniform interpolation produced output image with highest quality among all methods tested.
Primary Item Type:
Final Year Project
Language:
English
Subject Keywords:
grayscale image ; super-resolution ; sensors
First presented to the public:
6/1/2014
Original Publication Date:
1/16/2020
Previously Published By:
Universiti Sains Malaysia
Place Of Publication:
School of Electrical & Electronic Engineering
Citation:
Extents:
Number of Pages - 79
License Grantor / Date Granted:
  / ( View License )
Date Deposited
2020-01-16 15:54:19.883
Submitter:
Nor Hayati Ismail

All Versions

Thumbnail Name Version Created Date
Evaluation of super-resolution methods for grayscale image/Cheah Tatt Wei1 2020-01-16 15:54:19.883