Description
Programming section
Programming section:
Q1: Hough Transform for Line Detection
Q2: Fisher Linear Discriminant
Q1: Hough Transform for Line Detection
The goal of Q1 is to segment a A4 paper in the input image. A 4-steps
framework is provided in the skeleton code.
Framework for Q1
Step 1: Pre-processing.
Perform denoising on the
input image. Gaussian filter is
suggested in the skeleton
code. You can apply any other
filter.
Step 2: Edge extraction. You
can use the edge extraction
function that you developed
in assignment 1, or simply use
the built-in function edge().
CSIT 5410 Recognition Systems Assignment 2 Introduction
TA: Colin TSANG sctsangab@connect.ust./ 10
Framework for Q1
Step 3: Hough transform. Use
Hough transform to find four
straight lines that segment the
A4 paper. You CANNOT use
the built-in functions related
to Hough transform.
Step 4: Output. Plot the four
sides of the A4 paper in green
color.
Q2: FLD
The goal of Q2 is to develop a function for FLD.
Input:
class1 samples: Containing N1 data points in dimension N.
class2 samples: Containing N2 data points in dimension N.
input sample: Containing one data point in dimension N.
Output:
w: weight vector.
class: the decision of FLD (either 1 or 2).
Framework for Q2
Step 1: Obtain N1, N2, and N from the input.
Step 2: Compute w = S
−1
w (µa − µi). In this assignment, you do not
need to consider the cases where the matrix Sw is badly scaled,
singular or nearly singular.
Step 3: Compute the separation point for decision making by
1
2w
T
(µa + µi). Then output the decision for input sample.
