51st: LBP Interpretation _OPENCV

From: LBP (local binary mode) partial two value mode _ month to see the sunrise _ Sina Blog
Http://blog.sina.com.cn/s/blog_6e93dc1901010r15.html
LBP (local binary mode) partial two value mode (2012-03-15-10:38:17) Reprint Tags: it classification: Learning

LBP (local Binary patterns, locally two-valued mode) is an operator that describes the local texture characteristics of an image; Obviously, its function is to extract features, and the extracted features are texture features of the image, and are local texture features.

The original LBP operator is defined as the 3*3 window, with the window center pixel as the threshold, compares the gray values of the adjacent 8 pixels, if the surrounding pixel value is greater than the center pixel value, the pixel position is marked as 1, otherwise 0. Thus, the 8 points in the 3*3 field can produce a 8bit unsigned number, that is, the LBP value of the window, and this value is used to reflect the texture information for the region.

After the original LBP was presented, the researchers continually proposed various improvements and optimizations to obtain the LBP operator with P sampling points in the circular region with radius r, LBP uniform mode, LBP rotation invariant mode, LBP equivalence model, etc., please refer T. Ojala's article, published in IEEE Tpami in 2002, "Multiresolution Gray-scale and rotation invariant texture-classification with local Binary Patterns ".

It is obvious that the extracted LBP operator can get a LBP "code" at each pixel point, then, after extracting its original LBP operator, the original LBP feature is still "a picture". However, here we have converted objects from pictures (which can be understood as the measurement features of objects in the original measurement space) to two characteristics, which is what we normally call "features." However, this so-called "feature" can not be directly used in discriminant analysis. Because, from the above analysis, we can see that this "feature" is closely related to position information. Directly to two pictures extracted this "feature", and discriminant analysis, the "position is not aligned" and produced a lot of error. Later, the researchers found that a picture could be divided into several subregions, the LBP feature was extracted for each pixel point in each subregion, and then a statistical histogram of LBP features was established in each sub region. As a result, each subregion can be described with a statistical histogram, and the whole picture is made up of several statistical histograms; for example: a picture 100*100 pixel size, divided into 10*10=100, each of which is 10*10 pixels in size The LBP feature is extracted from each pixel in each sub region, and then a statistical histogram is established; In this way, this picture has the 10*10 sub area, also has the 10*10 statistical histogram, uses this 10*10 statistical histogram, can describe this picture. After that, we can judge the similarity of two images by using various similarity measurement functions.

At present, the LBP local texture extraction operator has been successfully applied to fingerprint identification, character recognition, face recognition, license plate recognition and other fields.