This article may be too technical for most readers to understand.June 2013) (Learn how and when to remove this template message)(
The core PatchMatch algorithm quickly finds correspondences between small square regions (or patches) of an image. The algorithm can be used in various applications such as object removal from images, reshuffling or moving contents of images, or retargeting or changing aspect ratios of images, optical flow estimation, or stereo correspondence.
The goal of the algorithm is to find the patch correspondence by defining a nearest-neighbor field (NNF) as a function of offsets, which is over all possible matches of patch (location of patch centers) in image A, for some distance function of two patches . So, for a given patch coordinate in image and its corresponding nearest neighbor in image , is simply . However, if we search for every point in image , the work will be too hard to complete. So the following algorithm is done in a randomized approach in order to accelerate the calculation speed. The algorithm has three main components. Initially, the nearest-neighbor field is filled with either random offsets or some prior information. Next, an iterative update process is applied to the NNF, in which good patch offsets are propagated to adjacent pixels, followed by random search in the neighborhood of the best offset found so far. Independent of these three components, the algorithm also use a coarse-to-fine approach by building an image pyramid to obtain the better result.
When initializing with random offsets, we use independent uniform samples across the full range of image . This algorithm avoids using an initial guess from the previous level of the pyramid because in this way the algorithm can avoid being trapped in local minima[clarification needed].
After initialization, the algorithm attempted to perform iterative process of improving the . The iterations examine the offsets in scan order (from left to right, top to bottom), and each undergoes propagation followed by random search.
We attempt to improve using the known offsets of and , assuming that the patch offsets are likely to be the same. That is, the algorithm will take new value for to be . So if has a correct mapping and is in a coherent region , then all of below and to the right of will be filled with the correct mapping. Alternatively, on even iterations, the algorithm search for different direction, fill the new value to be .
Let , we attempt to improve by testing a sequence of candidate offsets at an exponentially decreasing distance from
where is a uniform random in , is a large window search radius which will be set to maximum picture size, and is a fixed ratio often assigned as 1/2. This part of the algorithm allows the to jump out of local minimum through random process.
The often used halting criteria is set the iteration times to be about 4~5. Even with low iteration, the algorithm works well.
This section has multiple issues. Please help improve it or discuss these issues on the talk page. (Learn how and when to remove these template messages)(Learn how and when to remove this template message)
- Connelly Barnes, Eli Shechtman, Adam Finkelstein, Dan B Goldman(2009), PatchMatch: A Randomized Correspondence Algorithm for Structural Image Editing