L4.Image Coding Flashcards
How does frequency-domain processing make encoding of images more efficient?
- transform image into freq-domain
- encode only features that can be perceived by Eye
- reverse transform at receiver
Step 1 of JPEG image compression
- image 640x480 RGB pixels
- convert each to Y and ((I,Q) or (U,V))
- reduce I,Q to 2 320x240 matrices (averaging 2x2)
- consider Y,I,Q divided in 8-bit tiles
Step 2 of JPEG image compression
- DCT2 to each 8x8 tile
- (Fx,Fy,DCT) with small high frequency components
Step 3 of JPEG image compression
- Quantize DCT coeffs with table
- > divide
- > round to int
Step 4 of JPEG image compression
Encode differences in DC-DC components between tiles
Step 5 of JPEG image compression
- zig zag
- sum successive 0s (run length code )
Is RGB ro YIQ better for processing?
YIQ
DIfference between JPEG, MPEG and ZIP compression?
Jpeg, Mpeg are lossy
By what factor doe Jpeg compress images?
10x
By what factor does Mpeg compress videos?
100x
What choise decides the quality/compression trade-off?
Choice of Coefficient quentization matrix
What process of compressing increases sensitivity to bi-errors?
Huffman coding
Where is huffman coding used?
image compression
MP3 music encoding
Disadvantage of Huffman codes?
Sensitivity to bit error
What happens when a huffman code is affected by a bit error?
Variable length means if on symbol is affected all the ones following it are fucked
What does motion compensation mean in the context of encoding video?
- find similarity between parts of successive images
- send info where it is strong
- encode any jpeg differences
what 3 types of frame are there in mpeg?
I
P
B
What is an I-frame?
An Image encoded as JPEG
How does P-frame encode positions of moving blocks?
Predict from previous I and P frames
How does B-frame encode positions of moving blocks?
Estimate from both previous I&P and next
