25-08-2012, 05:11 PM
Principles of Image Compression
[attachment=32777]
Overview
• Image Compression is the Image Data
Elaboration branch dedicated to the image data
representation
• It analyzes the techniques allowing to reduce the
amount of data to describe the information
content of the image
Why Image Compression (1)?
• Images are usually matrices.
• Colour images are composed by 3 matrices of values.
• In a 24 bit per pixel (bpp) representation each color (e.g. Red,
Green, Blue) is represented with an unsigned byte in the range
[0;255]
• The 3 Matrices can be filed separately or together.
• The bytes number to store an uncompressed image can be really
huge!
How to compare?
The performance of an image compression technique
must be evaluated considering three different aspects:
Compression efficiency (Compression Ratio/Factor,
bit per pixel bpp or bit rate);
Image quality (Distortion Measure);
Computational cost.
Huffman Coding
Probability model and symbol-to-codeword are combined
Input: sequence of symbols.
Order the symbols according to their probabilities.
Apply a contraction to the two symbols with the smaller probabilities.
Repeat the previous step until the final set has only one member.
Construction of a binary tree:
The codeword for each symbol is obtained traversing the binary
tree from its root to the leaf corresponding to the symbol.
[attachment=32777]
Overview
• Image Compression is the Image Data
Elaboration branch dedicated to the image data
representation
• It analyzes the techniques allowing to reduce the
amount of data to describe the information
content of the image
Why Image Compression (1)?
• Images are usually matrices.
• Colour images are composed by 3 matrices of values.
• In a 24 bit per pixel (bpp) representation each color (e.g. Red,
Green, Blue) is represented with an unsigned byte in the range
[0;255]
• The 3 Matrices can be filed separately or together.
• The bytes number to store an uncompressed image can be really
huge!
How to compare?
The performance of an image compression technique
must be evaluated considering three different aspects:
Compression efficiency (Compression Ratio/Factor,
bit per pixel bpp or bit rate);
Image quality (Distortion Measure);
Computational cost.
Huffman Coding
Probability model and symbol-to-codeword are combined
Input: sequence of symbols.
Order the symbols according to their probabilities.
Apply a contraction to the two symbols with the smaller probabilities.
Repeat the previous step until the final set has only one member.
Construction of a binary tree:
The codeword for each symbol is obtained traversing the binary
tree from its root to the leaf corresponding to the symbol.