Lecture Notes - Introduction to Lossy Image Compression

Question 1

What is lossy image compression?

Question 2

What’s entropy?

Question 3

What’s entropy rate?

Question 4

Give the different combinations for the 4 subbands of a Haar transform and name the subbands

Question 5

Explain the concept of Multi-Level Haar. What’s the implication of the Multi-Level Haar are you increase the levels?

Question 6

After quantising the Haar transform, the huffman code can be used. Is it efficient or inefficient at reducing the entropy?

Answer 6

1. Inefficient
2. After quantising the Haar transform, the symbol ‘0’ is very frequent, with a probability much bigger than 0.5

e.g consider a black image. Huffman code will see it as each pixel bin with 1 bit and makes it harder to decode.

Question 7

In practical encoding, which coding scheme or rather combination gives near ideal entropy coding.

Answer 7

1. Pre-process the quantised transform by RLC
2. Then use Huffman.

Question 8

Why use the Haar transform?

Answer 8

The HT can be used to decrease the energy of the signal (or overall entropy) and thus achieve much better compression.