-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathindex.html
43 lines (41 loc) · 1.59 KB
/
index.html
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
<!DOCTYPE html>
<html>
<head>
<meta charset="UTF8">
<title>IVP Lab</title>
<style>
body{
background-color: #faf2e4;
margin: 0 10%;
font-family: sans-serif;
}
h1 {
text-align: center;
font-family: serif;
font-weight: normal;
text-transform: uppercase;
margin-top: 30px;
}
h2 {
color: #d1633c;
font-size: 1em;
}
h3 {
text-align: center;
font-family: monospace;
font-size 0.02em;
border-bottom: 1px solid #57b1dc;
}
</style>
</head>
<body>
<h1>Image & Video Processing Laboratory</h1>
<h3>Using computers to provide better mutlimedia experience</h3>
<h2>Image & Video Quality Assessment</h2>
<p>Comming up with objective quality metrics that can mimic human judgements improves the automatic parameter optimization of compression and resotration methods.</p>
<h2>Classic Compression & Coding</h2>
<p>We are adapting the existing compression standards for emerging type of media, such as <em>screen content images</em>.</p>
<h2>Neural Data Compression</h2>
<p>In an encoder-decoder CNN, the middle feature vector can be used to represent an image, given the network is trained to map an image to the very same input. That is, if you use the same input image as the ground-truth when training an encoder-decoder convolutional neural network, the encoder learns to compactly represent the input image with a feature vector of much lower dimensions (acting as a <em>compressor</em>), and the decoder learns to map the feature vecor (analagous to JPEG's <em>bitstream</em>) to the raw image (acting as the <em>decoder</em>).</p>
</body>
</html>