In digital image processing, the absolute difference number ( SAD ) is a measure of the similarity between image blocks. This is calculated by taking the absolute difference between each pixel in the original block and the corresponding pixels in the blocks used for comparison. These differences are summed to create a simple metric of the similarity of blocks, the norm L 1 of the difference image or Manhattan distance between two image blocks.
The number of absolute differences can be used for various purposes, such as object recognition, disparity map creation for stereo images, and motion estimation for video compression.
Video Sum of absolute differences
Contoh
This example uses an absolute amount of difference to identify which part of the search image is most similar to the template image. In this example, template images are 3 by 3 pixels, while search images are 3 by 5 pixels. Each pixel is represented by an integer from 0 to 9.
Image Search Templates  2 5 5 2 7 5 8 6  4 0 7 1 7 4 2 7  7 5 9 8 4 6 8 5
There are exactly three unique locations in the search image where the template might match: the left side of the image, the image center, and the right side of the image. To calculate the SAD value, the absolute value of the difference between each of the corresponding pixel pairs is used: the difference between 2 and 2 is 0, 4 and 1 are 3, 7 and 8 are 1, and so on.
Calculating the value of the absolute difference for each pixel, for three possible template locations, provides the following:
Left Right Center 0 2 0 5 0 3 3 3 1 3 7 3 3 4 5 0 2 0 1 1 3 3 1 1 1 3 4
For each of the three patches of this image, 9 absolute differences are added together, giving SAD values ​​of 20, 25, and 17, respectively. From this SAD value, it can be stated that the right side of the search image is the most similar to the template image, since it has the lowest absolute difference amount compared to the other two locations.
Maps Sum of absolute differences
Comparison with other metrics
Introduction to objects
The number of absolute differences provides a simple way to automate the search for objects within the image, but may not be reliable due to the effects of contextual factors such as changes in lighting, color, direction, size, or shape. SAD can be used in conjunction with other object recognition methods, such as edge detection, to improve the reliability of results.
Video compression
SAD is a very fast metric because of its simplicity; this is the most effective simple metric that takes into account every pixel in a block. It is therefore very effective for the search for wide movement of many different blocks. SAD is also easy to parallelize as it analyzes each pixel separately, making it easy to implement with instructions like ARM NEON or x86 SSE2. For example, SSE has packed the number of absolute difference instructions (PSADBW) specific to this purpose. Once a candidate block is found, the final refinement of the motion estimation process is often accomplished with slower but more accurate metrics, which better take into account human perception. These include the amount of absolute changed difference (SATD), sum of squares difference (SSD), and optimization of the rate of distortion.
See also
- Computer stereo vision
- Hadamard changed
- Motion compensation
- Approximate motion
- Introduction to objects (computer vision)
- Optimization of the distortion rate
References
-
E. G. Richardson, Iain (2003). H.264 and MPEG-4 Video Compression: Video Coding for Next Generation Multimedia . Chichester: John Wiley & amp; Sons Ltd.
Source of the article : Wikipedia
0 Comments