What are Different Methods for Searching for Motion Vectors?

Motion vectors are an essential component of video compression techniques. They allow for the efficient encoding of video data by identifying areas of the video that have moved between frames and encoding them separately. There are various methods for searching for motion vectors, each with its strengths and weaknesses. In this article, we will discuss the different methods for searching for motion vectors.

1. Full search method

The full search method is the simplest and most straightforward method for searching for motion vectors. It involves comparing each pixel block in the current frame with every block in the reference frame, in a specified search range. The search range can be defined as the distance in pixels from the current block to the reference block. This method is time-consuming as it requires a large amount of computational power to compare each pixel block with every block in the reference frame. However, it is accurate as it considers every possible block in the reference frame.

2. Three-step search method

The three-step search method is a faster version of the full search method. It involves comparing the current block with only a few blocks in the reference frame, rather than every block. The search pattern is usually a diamond pattern, starting at the center of the search range and moving outward in three steps. This method is faster than the full search method, but it may not always find the optimal motion vector.

3. New Three-step search method

The new three-step search method is an improved version of the three-step search method. It involves dividing the search range into four quadrants and searching each quadrant in a three-step search pattern. This method is faster and more accurate than the three-step search method, but it still requires a considerable amount of computational power.

4. Logarithmic search method

The logarithmic search method is a faster version of the full search method. It involves dividing the search range into smaller sub-blocks and searching each sub-block in a binary search pattern. This method is faster than the full search method, but it may not always find the optimal motion vector.

5. Diamond search method

The diamond search method is an improvement over the three-step search method. It involves comparing the current block with a set of blocks in the reference frame, arranged in a diamond pattern. The search range is gradually reduced until the optimal motion vector is found. This method is faster and more accurate than the three-step search method.

6. Hexagon search method

The hexagon search method is an improvement over the diamond search method. It involves comparing the current block with a set of blocks in the reference frame, arranged in a hexagon pattern. The search range is gradually reduced until the optimal motion vector is found. This method is faster and more accurate than the diamond search method.

7. Uneven multi-hexagon search method

The uneven multi-hexagon search method is an improvement over the hexagon search method. It involves dividing the search range into smaller sub-blocks and searching each sub-block in a hexagon pattern. This method is faster and more accurate than the hexagon search method.

Conclusion

In conclusion, there are various methods for searching for motion vectors, each with its strengths and weaknesses. The choice of method depends on the required level of accuracy and computational resources available. Full search method is the most accurate but also the most time-consuming, while the diamond search and hexagon search methods strike a good balance between speed and accuracy.