Video, Motion and MPEG

Question 1

Describe in bullet points, the process of Three-Steps Block Matching

Answer 1

1.Start with search location at center of image.

2. Set step size S=4 and search parameter p=7
3. Search 8 locations +/- S pixels around location (0,0) and the location (0,0)
4. Pick amount of the 9 locations searched, the one with minimum cost function.
5. Set the new search origin to the above picked location.
6. Set the new step size as S=S/2
7. Repeat the search procedure until S=1

Question 2

What are the pros and cons of Gradient-Based motion estimation with respect to Full Search Block Matching? In particular, discuss how they are affected by the aperture problem.

Answer 2

Advantages:
1. A sub-pixelic accuracy can be obtained with a few iterations. No need to make the full (2w+1)^2 evaluations of the full search BM.

Disadvantage:
1. The Taylor Series expansion is only valid over small distances.

Question 3

Explain where motion estimation is likely to fail or succeed in this frame.

Answer 3

1. Fast motion (size of motion bigger than the search window)
2. occlusions (object not visible in the other frame)
3. non-rigid objects (hair, flames)
4. motion blur (object is moving too fast and appears blurred)

Question 4

What is, in the context of video compression, a Group of Pictures? Name its different components.

Answer 4

A group of pictures, or GOP structure, specifies the order in which these frames types are arranged in a batch of frames.

Different components:
1. B = bi-directional frames
2. P = predictive frames
3. I = intra-coded frames

Question 5

Explain in bullet points the process of Gradient-Based Iterations for motion estimation.

Answer 5

1. Precomputing the gradients using for instance simple forward difference as approximation.

Question 6

What are the pros and cons of this Gradient-Based method over the Full Search Block Matching?

Answer 6

Advantages:
1. A sub-pixelic accuracy can be obtained with a few iterations. No need to make the full (2w+1)^2 evaluations of the full search BM.

Disadvantage:
1. The Taylor Series expansion is only valid over small distances.

Question 7

Explain how motion compensation is used to encode B-frames.

Answer 7

1. Have decoded forward frame In+j and backward frame In.
2. To reconstruct In, have 2 DFDs. (DFDb DFDf)
3. Keep the vector that gives the best prediction.

Question 8

Explain what a GOP in the context of MPEG and specify the types of frames in MPEG.

Answer 8

A group of pictures, or GOP structure, specifies the order in which these frames types are arranged in a batch of frames.

1. Bi-directional frames.
2. Predictive frames
3. Intracoded frames

Question 9

How could the GOP structure be optimised for the following?

(a) video editing application, where random access time is critical for interactivity
(b) video surveillance (static camera)
(c) video streaming.

Question 10

Name and explain 4 types of problems that can affect motion estimation.

Answer 10

1. Occlusion- object is not visible in the other frame.
2. Non-rigid objects - hair, flames.
3. Fast motion - The size of the motion bigger than the search motion.
4. Motion blur - The object is moving too fast and appears blurred.

Question 11

Why is full search block based matching?

Answer 11

Too computationally heavy

Question 12

What is gradient-based motion?

Answer 12

Gradient-Based motion estimation a class of algos that allows us to efficiently find the optimal a displacement, provided that the displacement is small.

Question 13

Define parameters ‘block size’ and ‘search radius’. Describe how the choice of these parameters affects the performance of block matching.

Question 14

State how the use of multiresolution approach improves the computational efficienecy of block matching compared with a conventional exhaustive search.