Assignment 4
Course Outcome No.: CO.CS311.4
Name: [Your Name]
Roll No: [Your Roll No]
Branch: [Your Branch]
Semester: [Your Semester]
1. What do you mean by Quantization? Describe the quantization problem with the help of an
example in detail.
Quantization is the process of mapping a large set of input values to a smaller set-often in digital
signal processing and data compression. It involves approximating a continuous range of values
with a finite set. This process introduces an error called quantization error.
Example: Consider analog-to-digital conversion. Suppose an analog signal ranges from 0 to 5V and
we need to digitize it using 8 levels (3 bits). Each level will cover a range of 5V / 8 = 0.625V. So, any
voltage between 0V and 0.625V is represented by 000, between 0.625V and 1.25V by 001, and so
on. A voltage of 1.1V would be quantized to 001, creating a small error. This simplification makes
storage and processing easier but introduces distortion.
2. Differentiate between uniform and non-uniform quantization.
Feature | Uniform Quantization | Non-Uniform Quantization
--------------------------|----------------------------------------|--------------------------------------------------
Step size | Constant across the range | Varies depending on the signal distribution
Complexity | Simple to implement | More complex
Application | Suitable for signals with uniform distribution | Better for signals with
non-uniform characteristics
Performance | Higher error for high-dynamic signals | Minimizes error by focusing on
frequent values
Uniform quantization is suitable for general-purpose applications, while non-uniform quantization is
better for voice and audio signals.
3. Describe Adaptive Quantization in detail and how it is different from uniform Quantization
technique.
Adaptive quantization is a dynamic form of quantization where the step size changes according to
the statistical properties of the signal. It adapts to signal variations over time. For example, during
silent periods in audio, fewer levels may be used, whereas more levels are used in active regions.
Difference from uniform quantization:
Uniform quantization uses fixed intervals, regardless of signal changes. Adaptive quantization
adjusts intervals based on real-time analysis. Hence, adaptive quantization provides better
performance in scenarios where signal properties change over time, like speech and video coding.
4. What are the various distortion criteria?
Distortion criteria refer to metrics used to measure the quality loss due to quantization. The main
criteria include:
- Mean Squared Error (MSE): Average of the square of the difference between original and
quantized values.
- Signal-to-Noise Ratio (SNR): Ratio of signal power to the power of quantization noise.
- Peak Signal-to-Noise Ratio (PSNR): Widely used in image compression, focuses on peak value of
signal.
- Mean Absolute Error (MAE): Average of absolute differences.
- Maximum Error: Largest single error between input and output values.
These metrics help in evaluating and optimizing quantizer performance.
5. What is entropy coded quantization?
Entropy coded quantization combines quantization with entropy coding techniques like Huffman or
arithmetic coding. After quantizing the signal, the quantized output symbols are encoded using
variable-length codes based on their probabilities.
Advantage: It reduces the average number of bits required to represent data. Frequent symbols get
shorter codes, while rare symbols get longer codes. This technique is widely used in standards like
JPEG and MPEG.
