What is a digital image?
? Digital image: an image can be defined as a two-dimensional function, where sum is the spatial (plane) coordinate, and the amplitude at any pair of spatial coordinates (,) is called the intensity or gray level of the image at that point. When the gray values of, and are finite discrete values, we call the image a digital image. A digital image consists of a limited number of elements, each of which has its specific position and amplitude. These elements are called drawing elements, image elements or pixels.
? This is because human perception is limited to the visible band of the electromagnetic spectrum, and imaging machines can cover almost the entire electromagnetic spectrum from gamma rays to radio waves. Digital image processing can deal with those image sources that humans are not used to.
? AI (Artificial Intelligence) is mainly divided into three parts: perception, understanding and decision-making. This kind of understanding is called image analysis (or image understanding) in image processing and computer vision. Professor Li Feifei, director of the Artificial Intelligence Laboratory (SAIL) of Stanford University, is very famous in the world. Understanding means understanding the deep meaning behind the image. The ultimate goal is to look at an old photo like a human being, which may make you cry (it contains a lot of information). Now, the achievements of Li Feifei Group can understand the relationship between objects. As shown in the figure below:
? Image processing is also divided into three levels: low-level processing, intermediate processing and advanced processing.
? Low-level processing: mainly do some simple operations on the picture, such as reducing noise, enhancing contrast and sharpening the image. Filtering can be used to reduce noise. The principle of image enhancement is to process a given image to make the result more convenient for subsequent operation and application than the source image, mainly to solve the problem of low contrast caused by the small gray range of the image, and to enlarge the gray level of the output image to a specified degree and make the details in the image clearer. Sharpen image features, such as edges, boundaries, contrast, etc. , so that the image can get better display effect or be more convenient for image analysis.
? Intermediate processing: it involves many tasks, such as dividing an image into different regions or targets. This is what the image segmentation field does to make it better recognized and classified, which can also be called target detection. Speaking of which, let's feel the result:
? Advanced processing: that is, understanding the above-mentioned images. Why is it so hard to understand? Because it is difficult for human beings to do this, just like in the eyes of ten thousand readers, there are ten thousand Hamlets. Although the field of digital image processing is based on mathematics and probability formulas, human intuition and analysis will play a central role in choosing one technology instead of another. In fact, the whole scientific field is like this.
? The concept of digital image processing existed as early as the 1920s, but the fundamental reason for its rapid development recently is that digital images need very large storage and computing power, so the development of digital image processing field must depend on the development of digital computers and related supporting technologies such as data storage, display and transmission. The improvement of computer computing power is also the fundamental reason for the development of AI. Many technologies were invented twenty years ago, and these contemporary scholars have made their effects even more amazing.
? Like the moon above, all information is hidden in pixels, but how to extract effective information? Give an example that is easier to understand:
? Reconstruct the original image through the diffuse light and shadow of the wall.
? The original image is on the left, the diffuse reflection image is in the middle, and the image reconstructed from the diffuse reflection image is on the right.
Name of article: Computational periscope with sequential digital camera
Article link:/articles/s 41586-018-0868-6
? Computer methods are used to enhance contrast, or to encode gray levels into colors, so as to facilitate the interpretation of X-ray images and other images in the fields of industry, medicine and biological sciences. Image processing technology has also been successfully applied to astronomy, archaeology, biology, nuclear medicine and law enforcement. Not very well), national defense and industrial fields.
? Having said so much about digital image processing, where does the image come from? Which is imaging technology. There are mainly gamma-ray imaging, X-ray imaging, angiography, ultraviolet band imaging, visible and infrared band imaging, microwave band imaging, acoustic wave imaging and so on. There is so much knowledge about design that I don't understand it. If you want to know, you can use Baidu keywords yourself.
? To put it bluntly, an image is a bunch of numbers, which can of course be generated directly out of thin air by a computer. Computer graphics is designed here, and combined with image processing, we get another field: three-dimensional modeling.