hjr-Digital Image Processing

Digital Image processing main content Digital Image representation grayscale color classification image Enhanced Grayscale Transform histogram equalization regulated image smoothing noise template convolution convolution neighborhood mean median filter image average image sharpening gradient operator Laplace operator sharpening non-sharpening filter high frequency enhanced filtering pseudo-color processing homomorphic Filtering

Digital Image processing the main content image acquisition representation and performance: image conversion, digital Image printing image restoration: Improve image quality (degradation cause known) image enhancement: Improve image quality (degradation cause known) image segmentation: Computers distinguish different objects in an image Image analysis: Image feature extraction, classification, recognition, understanding image reconstruction: According to the input data, transform the image Image compression coding: Compress large data volume image, easy to store and transfer Digital Image representation

Using a MXN matrix, the upper-left corner (0,0), the lower-right corner m-1,n-1, represents the image, and each element of the matrix is a pixel. Grayscale

Grayscale value 0-255, indicating from dark to bright, with different gray values to the sampling of successive values. Color

Each pixel, with RGB three values, three values range is [0-255], a total of 255*255*255 may, can represent a lot of colors, classification vectors: A formula to describe an image, small amount of data and scale the image is not distorted, difficult to make color-rich images. Bitmap: Represented by a pixel matrix, rich in color and large in data volume. Image Enhancement Grayscale Transformation

The gray range ranges from [A, b] piecewise linear or nonlinear transformations to [c,d]. Histogram

The gray-level function, the result of the frequency statistic of all shades, is independent of the position.
, which reflects the number or frequency of each gray level in the image. Equalization of

The histogram of the distribution concentration, the image shape unchanged, evenly distributed, enhance the contrast of the image. regulation of

The histogram changes according to the set rules. Image Smoothing Noise

In the image, the content should not appear in the frequency domain of high-frequency components. Template convolution

A 3x3 matrix for storing weighted coefficients, and convolution of the original image function convolution

Function A acts to function B, sets an interval, weights all the effects of a on B in this interval, and the weights generally decrease or increment with time. Neighborhood Average

The convolution template implements the average or weighted average of pixels adjacent to a pixel a, and then convolution the convolution operator with the original image. Median filter

Square window in the center of the pixel, all the gray values in the window are sorted from small to large, take the median value as the output result, and replace all the values in the window with the median value. Average Image

The matrix of multiple images of the same scene accumulates and then averages. Image Sharpening gradient operator

The image is the matrix on the x, Y axis, the gradient matrix is based on the biased x, and the matrix of the biased Y, gradient is the deviation of the guide, is high school slope, reaction image gray change situation, second-order differential is to find two times guide is called second-order deviation, the change rate of reaction change. And the distance divided by the time equals the speed, the speed derivative equals the acceleration and the like. Laplace operator

Implements a 2x2 convolution matrix for 2-order biasing in the X and y directions. Sharpening

The matrix obtained from the convolution of an image f (x, y) and a gradient operator or a Laplace operator is a non-sharpening image filter after sharpening.

Matrix subtraction High frequency enhanced filter of original image matrix minus sharpening

Original image matrix plus sharpening matrix pseudo-color processing (time domain) different shades of gray are represented by different colors (frequency domain) different frequencies with different colors to represent the homomorphic filter

Automatically adjusts image quality according to lighting changes