=========================preview======================
(ELEC360)quiz01a.pdf
Back to ELEC360 Login to download
======================================================
Hong Kong University of Science & Technology Department of Electrical & Electronic Engineering
ELEC 360: Digital Media & Multimedia Applications
Fall 2002
Name: _______________________________ Student ID: ___________________________ E-mail: ______________________________
Quiz #1A (October 15, 2002)
1.
Consider the 24-bit RGB pixel representation (8 bits for each R, G and B component). White is <255,255,255> whereas black is <0,0,0>. What is the RGB value of the complementary color of <123,125,126>?

2.
Non-linear transformation is commonly applied to the audio signal before it is uniformly quantized in the amplitude domain. In the following figure, the horizontal axis represents the input analog signal (Vin) and the vertical axis represents the output analog signal (Vout). The slanted dashed line indicates that Vin == Vout. Suggest a non-linear transformation from Vin to Vout such that the louder input signal will ultimately bear less amplitude quantization noise.


Vout Vout = Vin

0 Vin
3.
Name one of the masking properties of human hearing that is used by the perceptual audio coding scheme. Briefly describe how it works.

4.
Describe how you would create the digital reverberation effect.

5.
In the linear predictive coding (LPC) audio compression scheme, which one of the following items is NOT transmitted to the decoder?

a.
order of the prediction

b.
prediction coefficients

c.
prediction residue

d.
code book



6.
How would you convert a 24-bit RGB pixel (8 bits for each R, G and B component) to a 16-bit pixel for uniform representation in the RGB color space?

7.
Consider the following 2x2 luminance pixel block on the left. We want to scale it up to the 3x4 pixel block. Apply linear interpolation to estimate the `missing pixels in on the right.

8.
In JPEG compression, each color component is divided into 8x8 pixel blocks. Each block undergoes forward DCT (Discrete Cosine Transform). It is then followed by DCT coefficient quantization. Suppose the following 8x8 block is taken from one of the quantized result. In order to represent the result in a more compact way, we go through the quantized coefficients and apply run-length encoding (RLE). Basically, we pair up the consecutive number of zero coefficients (called zero-run) with the terminating non-zero coefficient in the scan path. In JPEG, a zig-zag path is used. It starts from the lowest spatial frequency term and winds its way to the highest spatial frequency component. DC is not included because it is coded differently. Either use the JPEG zig-zag or use your own scanning method and write down all the pairs. For example, <number of zeros, value>1, <number of zeros, value>2, <number of zeros, value>3,..Remember to sketch the scanning path you take with a few arrows to indicate the scanning direction.

9.
The following chart is used in the spatial resolution testing. Circle the region from t