Language
Country/Area
No result found
The big world of tiny macroblocks: How to decode the entire video in 16x16 pixels?!
In the context of digital video compression, the block matching algorithm has become the core technology of motion estimation. The basic assumption of the algorithm is that in adjacent video frames, patterns of the same object and background will move within the frame over time to form new corresponding objects. Through this mode, the algorithm is able to capture temporal redundancy in the video sequence, thereby increasing the effectiveness of compression between frames. The blocks of the current frame are split into macroblocks and then compared with the corresponding areas of the previous frame, and the differences are used to find the best match.
"By predicting the motion vectors in the video, the amount of data required can be effectively reduced, which is crucial for video compression."
Block matching algorithms typically divide the current frame into 16x16 pixel macroblocks and compare these blocks with corresponding areas in adjacent frames. The result of this comparison generates a vector that describes the motion of the macroblock from one position to another. The motion vectors of all macroblocks make up the motion estimate for the frame. The search range for a match is determined by the "search parameter", which sets the number of surrounding pixels of the corresponding macroblock in the previous frame to be considered.
The main purpose of motion estimation is to obtain motion vectors that can effectively represent the motion from one 2D image to another. These motion vectors may involve global motion estimation of the entire image or motion in a specific area, such as a rectangular block or Blocks of any shape. In addition, these vectors can be matched and described by multiple models based on the motion characteristics of real video cameras, such as rotation or scaling."The combination of motion estimation and motion compensation is a key part of many current video coding standards, such as MPEG and H.261."
However, motion estimation is the most computationally intensive process in the entire video compression process, so a faster and more computationally efficient algorithm is needed. For example, comprehensive search can achieve the best matching results, but its computational cost is extremely high. Some more efficient algorithms, such as optimized hierarchical block matching (OHBM) and three-step search (TSS), are proposed to meet this computational requirement.
Evaluation Metrics
During the block matching process, it is very important to evaluate the matching degree of macroblocks, usually based on a cost function. For example, one of the most popular evaluation methods is the mean absolute error (MAD), but there are also various indicators such as the mean squared error (MSE).
“These cost functions not only affect the performance of the algorithm, but also directly affect the loss of final video quality.”
Various Algorithms
Many block matching algorithms have been proposed and developed since the mid-1980s. Here are some popular algorithms:
Exhaustive Search
This is the simplest but most computationally intensive algorithm, computing the cost function for all possible positions to find the best match. Although this approach provides the best signal-to-noise ratio, it also requires enormous computing resources.
Three Step Search
This is an early fast block matching algorithm that significantly reduces the amount of computation by gradually narrowing the search range.
Diamond Search
Algorithms using the diamond search pattern can find the global minimum match at a lower computational cost and provide a peak signal-to-noise ratio close to that achieved by a full search.
With the advancement of technology, more efficient algorithms such as Adaptive Root Pattern Search (ARPS) have begun to emerge, which can effectively use the motion of surrounding macroblocks to infer the motion of the current macroblock, thereby reducing the calculation time.
Finally, the development of block matching algorithms has become an important cornerstone of video coding technology. This not only improves the efficiency of video compression, but also paves the way for future image processing and transmission technologies. Therefore, can we imagine what the future video world will look like if block matching technology continues to advance?
