Image Enhancement and Image Restoration

Before analyzing images, you must improve the image quality. Generally, there are two ways to improve the image quality: image enhancement.And Image Restoration

 
Image Enhancement does not take into account the cause of image quality decline. It only highlights the features that interest the image, but degrades the features that are not needed. The purpose of this feature is to improve the readability of the image. Image enhancement methods can be divided into two types: the airspace method and the frequency domain method. The airspace method is mainly used to operate each pixel in the image. The frequency domain method is used to operate the image in a certain transform domain of the image, modify the transformed coefficient, such as the Fu Yi transform and DCT transform, and then perform inverse transformation to obtain the processed image.

Airspace method: Point processing (Image Grayscale transformation, histogram equalization, pseudo-color processing, etc)

Frequency Domain Method: high and low-pass Filtering

Image Enhancement Methods:

Grayscale conversion:

A slightly more complex method for gray-scale image transformation is histogram equalization. Histogram equalization is an important application of grayscale transformation. It is widely used in image enhancement. It is a histogram correction method based on the cumulative distribution function transformation, an image with a gray-level distribution with a uniform probability density can be generated, extending the dynamic range of pixels.

Histogram equalization:

Histogram equalization transforms the histogram of the original image into a uniform distribution, so as to increase the dynamic range of the gray value to enhance the image.

Image Smoothing:

Image Smoothing is mainly used to eliminate noise. Noise is not limited to the distortion and deformation that the human eye can see. Some noises can only be detected during image processing. Common noise of images includes supplementary noise, multiplication noise, and quantization noise. The noise in the image is often intertwined with the signal, especially the multiplication noise. If it is not smooth, the details of the image, such as the border contour and lines, will become blurred, how to smooth out noise and keep image details as much as possible is the main task of image smoothing.

In general, the image energy is mainly concentrated in its low-frequency part, the noise is mainly located in the high-frequency band, and the edge information of the image we want to extract is also concentrated in its high-frequency part. Therefore, the important thing is how to remove high-frequency interference while keeping edge information. To remove noise, it is necessary to smooth the image. Low-pass filtering can be used to remove high-frequency interference. Image Smoothing includes the airspace method and the frequency domain method. In the airspace method, the common method of image smoothing is mean filtering or mean filtering, it uses a sliding window with an odd number of points to slide the image. The gray value of the image pixel corresponding to the center of the window is replaced by the average value of the gray value of each point in the window, if the Sliding Window specifies the weight of each pixel in the window, that is, the coefficient of each pixel, in the process of obtaining the mean, this is called weighted mean filtering. For median filtering, the gray value of the corresponding pixel is replaced by the median value in the window. To simplify programming, you can define an N * n template array for average or median filtering. In addition, you need to note that the pixels at the four edges of the image can be left blank during image scanning using Windows; you can also use pixels with a gray value of "0" to expand the image edge.

Image sharpening:

Image Smoothing often blur the border and contour of an image. To reduce the impact of such adverse effects, the border and contour technology of the image must be used to make the image clearer. The goal of image regionalization is to make the image's edge, contour, and image details clearer. The root cause of smooth image blurring is that the image is subjected to average or integral operations, therefore, you can perform inverse operations (such as differential operations) to clear the image. Considering from the frequency domain, the essence of Image Blur is that its high-frequency components are degraded, so a high-pass filter can be used to make the image clear.

Sharpening is used to enhance the gray contrast. Because the edge and contour are both located in the gray mutation, the sharpening algorithm is implemented based on differentiation.

 
Unlike image restoration and enhancement, image restoration needs to understand the cause of image quality decline. First, we need to establish a "downgrading model" and then use this model to restore the original image.

 
Image Restoration Methods are as follows :(

Image Degradation: The distortion problem that occurs when the image obtained from the scene does not fully reflect the actual content of the scene .)

Cause of degradation: recovery method:

Noise Removal caused by adding noise System-caused system degradation Inverse Operation

  Image Restoration uses a prior knowledge of the degradation phenomenon to establish a mathematical model of the degradation phenomenon, and then reverse deduction based on the model to restore the original scene image. Therefore, image restoration can be understood as a reverse process of image quality reduction. Establishing a mathematical model of the reverse process of image restoration is the main task of image restoration. After the calculation of the mathematical model in the inverse process, it is difficult to restore the full scene image. Therefore, image restoration often requires a quality standard,
That is, to measure the degree close to the image of a fully real scene, or to determine whether the estimation of the original image has reached the optimal degree. Due to the large number of factors causing degradation and the different nature, the mathematical models established to describe the image degradation process are often diverse, and the quality standards for restoration are often different, therefore, image restoration is a complex mathematical process, with different methods and techniques.

What is the relationship between image restoration and image enhancement?

Image enhancement is a heuristic process. The purpose of this process is to enable the human eye system to observe the image content. "image enhancement" is a heuristic process, the purpose is to make the processed image easier for the human eye system to observe the image content.

(1) similarities: To improve the visual effect of a given image in a certain sense, some content is the same. (2) differences: A. Image Restoration can use a prior knowledge of degradation phenomena (degradation models) to reconstruct and restore degraded images. This technology needs to identify the cause of degradation, perform mathematical modeling, and then restore the image along the inverse process of image degradation. Image enhancement technology basically does not model the process of image degradation or degradation, and only pursues better visual effects to meet the needs. B. There are clearly defined quality standards for image restoration technologies, so that the best assessment can be provided for restored images. Image enhancement technology is a kind of psychological acceptance process. Generally, there is no objective and unified evaluation standard. C. Image enhancement is a heuristic process. The objective of the processed image is to help the human eye system observe the image content.

"Image Restoration" is based on a certain optimal principle, so that the restored image is the best approximation of the ideal image.