DCT, JPEG and Compression

Question 1

Name two reasons why JPEG uses YCbCr instead RGB

Answer 1

1. YCbCr seperates the brightness information (Y) from the colour information (Cb and Cr).
2. YCbCr enables retention of essential visual details while reducing data size.
3. Aligns well with the human eye’s differing sensitivities to luminance and chroma.

Question 2

Briefly explain how the DC and AC components are encoded in JPEG.

Answer 2

DC Coefficients

1. The DC coefficient exhibits different statistics from the remaining 63 AC coeffcients.
2. DC coeffs exhibit local correlations so differential coding is used.
3. Differential coding = difference between current DC coef and previous DC coef.

AC Coefficients

1. AC components (-1023 1023) are coded as pairs of symbols: (RL, size)(Value)

Question 3

What do contrast sensitivity graphs tell us about the human sensibility to spatial frequencies? How are they exploited in the JPEG standard?

Answer 3

1. Humans see mid-frequencies best; poor at very low/high ones.
2. JPEF exploits by keeping low/mid frequencies and discarding high frequencies.

Question 4

Why is the Run length encoding used in image processing? How is this used in JPEG?

Answer 4

1. To get the bit rate < 1, by encoding consecutive zero values.
2. Used in encoding the AC coefficients.

Question 5

What is differential coding and how is it used in JPEG?

Answer 5

1. Differential coding is used to find the difference between the previous and current DC coeffs.

Question 6

What are Run Length codes? How is this used in JPEG?

Answer 6

A coding scheme that only encodes consecutive zeros. (i.e. each run of consecutive zero samples in the vector is coded as a single symbol with an associated length)

-13 -5 1 0 0 0 0 0 0 -1

-13, -5, 1, 4x0

Question 7

After having the DCT coefficients on 8x8 blocks, explain how JPEG proceeds to get an encoded bitstream.

Answer 7

1. DCT coefficients are quantised.
2. Zig-zag scan, which gets high to low frequencies in the block.
3. DC coeffs -> differential coding.
4. AC coeffs -> RLE (due to lots of zeros.)

Question 8

How is the DCT related to JPEG?

Answer 8

The 8-point DCT is the basis of the JPEG standard, as well as several other standards.

Question 9

What does the EOB tell you?

Answer 9

1. End of Block
2. “no need for encoding” zeros afterwards

Question 10

Explain what ZRL is.

Answer 10

There are 16 zeros after.