Do some real-time image processing, license plate number recognition using the phone's camera, then no special many types of image processing
OpenCL (Open Computing Language): Open Computing Language, can be attached executed on the CPU or GPU of the host processor.
OpenCV library includes support for OpenCL and CUDA GPU architecture. OpenCV has a new unified data structure UMat, for the necessary and possible, is responsible for the transmission of data to the GPU.
OpenCv-- image blurring (mean, Gaussian, median, bilateral) - https://blog.csdn.net/qq_39861376/article/details/82109127
image processing algorithms used (C ++ / OPENCV) - https : // blog. csdn.net/thecentry/article/details/80698872
the OpenCV using Demo - https://github.com/muyishuangfeng/OpenCV
image smoothing - http://www.opencv.org.cn/opencvdoc/2.3.2/html/ doc / tutorials / imgproc / gausian_median_blur_bilateral_filter /
gausian_median_blur_bilateral_filter.html # median-filter smoothed, also known as fuzzy, it is a high frequency of use and simple image processing method. When the need to use a smoothing filter. The most commonly used filter is a linear filter, the output pixel value of the linear filter processing (ie g (i, j) ) is the input pixel value (ie f (i + k, j + l)) and weighted. Normalization block filter (Normalized Box Filter) blur, Gaussian filter (Gaussian Filter) GaussianBlur, median filter (Median Filter) medianBlur, bilateral filtering (Bilateral Filter) bilateralFilter.
One of the most simple image processing operation and the fuzzy operation common, one of the reasons for the operation of the time used to reduce the pre-noise to the image.
Is a linear Gaussian smoothing filter for eliminating Gaussian noise, a noise reduction process is widely used in image processing. Popular speaking, the Gaussian filter is to process the whole image of the weighted average value of each pixel, and the other by its own neighborhood pixel values obtained after a weighted average.
Specific operation Gaussian filter is: a template (or convolution mask) scanning each pixel in the image, the weighted average of gray value pixels in the neighborhood determined by the value of the template to replace the central pixel of the template.
OpenCV- With me Laplace operator of digital image processing, OpenCV, digital image processing of the Laplace operator. Laplace operator is the second derivative operation, which does not emphasize the grayscale values in successive portions in FIG emphasize grayscale pixel discrete portions simultaneously. Laplace operator and Sobel operator as part of the spatial filter sharpening operation.
OpenCV achieve grayscale image -> binarization -> the Canny edge detection and contrast (Contrast) and brightness (the Brightness) value adjusting
gray image: simply means that to make R, G, B value in the threshold value [0 , 255] to obtain the same value among, a total of 256 levels, in fact, can be seen as white and black colors in the case when the gradation of the two extremes, intermediate level 254 representing the degree of gradation (shallow depth) .
Image Binarization: simply means that the gradation value of the pixel on the image is (0,0,0) (black) or (255,255,255) (white), showing a clear image only black and white effect.
Canny operator: The aim is to find an optimal edge detection algorithm to satisfy three main criteria:
1. a low error rate: identifying as many edges, while minimizing the influence of noise generated;
2. High Location: identifying edges should be as close as possible the actual edge image;
3. minimum response: edges in the image can only be identified once, and there may not image noise is identified as an edge.
Canny edge detection step: Filter -> calculate the gradient magnitude and direction -> Non-maxima suppression (excluding non-edge pixel, the candidate edge retention)
Brightness and contrast adjustment: to begin with what point operation (point operator) this concept: The input pixel value (sometimes add some global information or parameters), to calculate the value of the output pixel. Such operators including brightness and contrast adjustment, color correction (colorCorrection) and conversion (transformations). The two most common operating point satisfies the following formula:
G (I, J) = A * F (I, J) + B, (A contrast control, b control the brightness, f (i, j) of the source image pixel, g (i, j) is the output image pixels)
The image consists of an array, a single channel of a monochrome image is two-dimensional matrix, like a checkerboard (matrix), the size of the digital board of each point represents the gradation level of image pixels, sketch-like black and white pixels is formed by the difference image; and a color image is a three-dimensional matrix formed by a combination of the three channels. Color images can be achieved by b, g, r = cv2.split ( image) isolated, or sub-channel to achieve separation of b, g, r = image array by cutting [:,:, 0], image [:,:, 1] , image [:,:, 2 ]; image combining three channels of a color image can call cv2.merge ([b, g, r ]) were combined channel to achieve.
Digital image processing is the nature of the operation grayscale matrix. Digital image processing means by means of digital image processing digital computer. Note that the digital image is of a limited number of elements, each element has a specific location and magnitude.
Geometric transformation of the image, a complete image of the chapter requires two separate geometric transformation algorithm, first of all, needs to have a spatial coordinate conversion, each pixel is described how to use it to move from an initial position to the end position; Secondly, a required interpolation algorithm gradation value of each pixel of an output image is completed.