Dissertation/Thesis Abstract

Computer Vision and Image Processing Toolbox for MATLAB
by Regmi, Krishna, M.S., Southern Illinois University at Edwardsville, 2015, 113; 1597577
Abstract (Summary)

Introduction: The CVIPtools software was developed for different image processing and computer vision operations with the code written in C language. These codes can be reused to execute the functions in MATLAB. To make the C code executable in MATLAB an intermediate function called a wrapper function has been written for each function. These functions are compiled to create dynamically linked libraries (dll) and these dll are used to create a toolbox along with CVIP-C functions and wrapper functions. Once toolbox is created, these functions are executed as built-in functions in MATLAB. The advantage of this is that the functions need not be compiled each time it is executed in different machine. So, the user does not need to know all the minute details of each function before executing them. Also, this toolbox is independent to other toolboxes; user does not need to install some other toolboxes to execute functions from this toolbox. A help file has been created for the convenience of the users regarding information about parameters each functions take.

Objectives: The primary objective of this research study was to write wrapper functions which are intermediate functions that facilitate the execution of CVIP-C functions and code in MATLAB. Once the wrapper functions were written, next step was to create a toolbox that eases the execution of the functions.

Methodology: The wrapper functions were written for the C-functions of CVIPtools software. These wrapper functions are MATLAB executable (MEX) functions. These functions facilitate the execution of C-functions in MATLAB. Visual C Compiler is used to compile the MEX files. Once compiled, the dynamically linked library files are created. These dll files are used to execute the functions. These are also used to create the toolbox. The toolbox once installed by user in his/her machine can execute the functions without necessity of compiling. A total of 99 functions have the wrapper functions written and are integrated in the toolbox.

Results: After the wrapper functions were written, the toolbox was created and the functions were executed in MATLAB. After execution, the output images produced were saved to the disk. These output images were compared with output obtained from execution of same function in CVIPtools and the results were analyzed. Among the 99 functions in toolbox, 66 functions produced exactly the same images. The 17 functions produced the same outputs in both platforms but the data-type of output images are FLOAT so to save it needed to be changed to BYTE. The other 12 functions produced outputs with slight discrepancies in some pixel values. The remaining four functions produced correct outputs but the operations were not available in CVIPtools for comparison; only the C code was available to write wrapper functions.

Conclusion: The research study supports that the C-code and C-functions can be executed in MATLAB by writing intermediate functions that act as a bridge between two platforms. The toolbox created during this research can be distributed to user community so that the image processing and computer vision operations can be accomplished using a single toolbox which is independent to other toolboxes.

Indexing (document details)
Advisor: Umbaugh, Scott E.
Commitee: Smith, Scott, Wang, Xin
School: Southern Illinois University at Edwardsville
Department: Electrical and Computer Engineering
School Location: United States -- Illinois
Source: MAI 55/01M(E), Masters Abstracts International
Subjects: Engineering
Publication Number: 1597577
ISBN: 978-1-339-00930-8
Copyright © 2020 ProQuest LLC. All rights reserved. Terms and Conditions Privacy Policy Cookie Policy