Determining Edges with a Sub-Pixel Accuracy
Step 1: Apply an edge detector (e.g., Sobel) to isolate the edges in the image.
Threshold the image leaving only the edges (i.e., areas with an
elevated spatial derivative)
Step 2: Determine the location of the
Image Restoration
Many image enhancement techniques are designed to compensate for the
effects of a known or estimated degradation process.
This is referred to as image restoration.
On of the most common deterministic spatial degradations is blurring.
Blu
Edge Detection: Laplacian Operator
0 1 0
1 4 1
0 1 0
Disadvantages:
The Laplacian operator is more sensitive to impulse (salt-andpepper) noise than actual edges in the image.
The results of the Laplacian operator are two times higher for
diagonal edg
Image Enhancement
Methods of evaluating image quality
enhancing low quality pictures
Techniques
o Gray Scale Transformation
o Histogram Modification
o Sharpening
o Filtering, smoothing, noise removal, averaging
Contrast:
Low
Good
High
Gray Scale Tra
Image Restoration
Dr. Kaaren May
Memorial University
June 2012
Overview
What is image restoration?
Laplacian approach to restoration
Explicit modeling of blur and noise
Inverse filtering
Wiener filtering
2
Image Restoration
You have an image that has bee
Filtering: Averaging
Assume we have access to
versions of the same image.
(
)
Calculate the average of the
(
versions on a pixel by pixel basis:
)
(
)
) consists of a gray level intensity value, (
Each pixel in (
).
additive noise, (
After averaging s
Structural Representation of Images
The segmentation process decomposes a digital image, D,
into n subsets or regions denoted by R1, R2, , Rn.
R1 R2 . Rn = D
Ri R j = 0
for
i j
Each pixel in Ri has a characteristic value associated with the
region, vi (
Mathematical Model of Image Formation Process (Camera
Model)
Forward camera model: converts the 3D world coordinates of a point in
space to distorted pixels in a computer
The transformation describing the image formation process can be
summarized as follo