Department of Computer Science Columbia University
Group Project: iris recognition
Instructor: Xiaofu He Fall 2019
Requirement
1) Implement the exact same design as Ma et al., 2003 paper (see the reference) but focus on Image Preprocessing, Feature Extraction, and Iris Matching only and use provided dataset.
2) Download the CASIA Iris Image Database (version 1.0) (CASIA-IrisV1) from the Coursework and unzip the dataset to the same directory as your Matlab or python scripts.
3) Three students per group (need to submit peer evaluation form). If you cant find group members, please let my TA know.
Experiment design
Database: 108 eyes, 7 iris images per eye, which were captured in two sessions (3 in the first session, 4 in the second session). All images are stored as BMP format with 320280 pixel size.
Experiment design: images from the first session will be used for training and images from the second session will be used for testing
Experimental results:
The Correct Recognition Rate (CRR) for the identification
mode (refer to Tables 3 & 10 of Mas paper)
Receiver Operating Characteristic (ROC) curve for the verification mode (refer to Table 4 and Fig. 13. of Mas paper)
Submission
Source codes
IrisRecognition.m/py: the main function, which will use all the following sub
functions:
IrisLocalization.m/py: detecting pupil and outer boundary of iris. You can choose other iris localization methods if they work better;
IrisNormalization.m/py: mapping the iris from Cartesian coordinates to polar coordinates;
ImageEnhancement.m/py: enhancing the normalized iris;
FeatureExtraction.m/py: filtering the iris and extracting features;
IrisMatching.m/py: using Fisher linear discriminant for dimension reduction and nearest center classifier for classification;
PerformanceEvaluation.m/py: calculating the CRR for the identification mode (CRR should be >=75%), which will output Tables 3 & 10 (refer to Mas paper); calculating ROC curve for verification mode, which will output Table 4 and Fig. 13 (using Bootstrap and calculating confidence interval are optional).
For each script, 1). explain the logic behind the script, e.g., what is the loop for? what is that piece of code for? 2). specify and explain ALL variables/parameters used in the script.
Extra bonus
calculating the CRR for the identification mode (CRR should be >=85%, and using Bootstrap and calculating confidence interval, you can get extra credit.
A readme file (UNI_UNI2_UNI3_IrisRecognition.README)
Explain the whole logic of your design.
Briefly discuss the limitation(s) of the current design.
How can you improve it? Peer evaluation form
Compress all files into a single zip file with UNI_UNI2_UNI3_IrisRecognition.zip/.rar as its name
All scripts must be runnable (we wont debug for you).
Do not submit the dataset.
Submit to the coursework, due on Oct 30th (11:59PM)
Peer Evaluation Form
Write the name of each of your group members in a separate column. For each person, check the box if this part is contributed by this group member. If this part is completed by multiple group members, please specify it in comments.
Part
Group member:
Group member:
Group member:
comments
IrisRecognition.m/py
IrisLocalization.m/py
IrisNormalization.m/py
ImageEnhancement.m/py
FeatureExtraction.m/py
IrisMatching.m/py
PerformanceEvaluation.m/py
Readme file
Reference
Ma et al., Personal Identification Based on Iris Texture Analysis, IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 25, NO. 12, DECEMBER 2003
Note_CASIA-IrisV1.pdf
Reviews
There are no reviews yet.