Overview of Digital Image Processing

Digital image processing is also called computer image processing.
The process of converting an image signal into a digital signal and processing it using a computer. Digital image processing first appeared in 1950s. At that time, electronic computers had developed to a certain level, and people began to use computers to process graphics and image information. Digital image processing, as a discipline, was developed in the early 1960s S. Early image processing aims to improve the quality of the image. It takes people as the object for the purpose of improving the human visual effect. In image processing, images with low quality are input, and images with improved quality are output. common image processing methods include image enhancement, restoration, encoding, and compression. The first successful application was the US Jet Propulsion Laboratory (JPL ). They used image processing techniques, such as geometric correction, gray-scale conversion, and noise removal, to process thousands of lunar photos sent back by the space detector's inner 7 in 1964, taking into account the influence of the sun position and the lunar environment, the computer successfully drew a map of the lunar surface, and achieved great success. Then, the detection ship sent back nearly 100,000 photos for more complex image processing, resulting in the acquisition of the lunar topographic map, color map and panoramic mosaic, achieved remarkable results, this has laid a solid foundation for the launch of the "Moon Landing" initiative and promoted the birth of the digital image processing discipline. Digital image processing technology has played a huge role in future aerospace space technologies, such as the exploration of Mars, Saturn and other planets. Another major achievement of digital image processing is the achievement of medicine. In 1972, the British EMI engineer housfield invented the X-ray computer tomography device for skull diagnosis, which we usually call computer tomograph ). The basic method of CT is to reconstruct a cross-section image based on the projection of a person's head section, which is called image reconstruction. In 1975, EMI successfully developed a whole-body ct device, and obtained clear and clear tomography images from various parts of the human body. In 1979, this harmless diagnosis technology won the Nobel Prize, indicating that it has made an epoch contribution to mankind. At the same time, image processing technology has been widely valued in many application fields and has made significant pioneering achievements, these fields include aerospace, biological and medical engineering, industrial inspection, robot vision, Public Security and Justice, military guidance, culture and art, make image processing a new discipline that is eye-catching and promising. With the in-depth development of image processing technology, since the middle of 1970s, with the rapid development of computer technology, artificial intelligence, and thinking science research, digital image processing has developed to a higher and deeper level. People have begun to study how to use computer systems to interpret images and make it possible for human visual systems to understand the external world. This is called image understanding or computer vision. Many countries, especially developed countries, have invested more manpower and material resources in this study and have made many important research achievements. The representative result is the visual computing theory proposed by Marr at the end of 1970s. This theory has become the dominant idea in the computer vision field for more than 10 years. Although Image Understanding has made great progress in the study of theoretical methods, it is a difficult research field and has many difficulties, because humans do not know much about their own visual processes, computer vision is a new field to be further explored.
The main research contents of digital image processing include the following: 1) image transformation is directly processed in the spatial domain due to the large size of the Image array, which involves a large amount of computing. Therefore, various image transformation methods, such as Fourier transformation, Fourier transformation, discrete cosine transformation, and other indirect processing technologies, are often used to convert the processing of spatial domains into transform domains, this not only reduces the amount of computing, but also achieves more effective processing (for example, Fourier transform can be used for digital filtering in the frequency domain ). At present, the emerging research of wavelet transform has good localization characteristics in the time domain and frequency domain, and it is also widely and effectively used in image processing. 2) image encoding compression technology can reduce the amount of data describing the image (that is, the number of BITs), in order to save image transmission, processing time and reduce the memory capacity occupied. Compression can be achieved without losing the truth, or under the allowable distortion conditions. Encoding is the most important method in compression technology. It is the earliest and mature technology in image processing technology. 3) image enhancement and Restoration The purpose of image enhancement and restoration is to improve image quality, such as noise removal and image clarity. Image Enhancement highlights the areas of interest in the image, regardless of image downgrading. For example, the enhancement of the image's high-frequency component can make the object contour clear and the details obvious. For example, the enhancement of the low-frequency component can reduce the noise impact in the image. Image Restoration requires a certain understanding of the cause of image quality reduction. Generally, we should establish a "Quality Reduction Model" based on the quality reduction process, and then adopt a certain filtering method, restore or recreate the original image. 4) image segmentation is one of the key technologies in digital image processing. Image segmentation is to extract meaningful features from an image. Its meaningful features include the edge and area of the image. This is the basis for further image recognition, analysis, and understanding. Although many Edge Extraction and region segmentation methods have been developed, there is no effective method that is applicable to various images. Therefore, the research on image segmentation is still in depth and is one of the hot topics in image processing. 5) image description is a prerequisite for image recognition and understanding. As the simplest binary image, the geometric characteristics of the object can be used to describe the characteristics of the object. The general image description method adopts two-dimensional shape description, which has two methods: boundary Description and region description. Two-dimensional texture features can be used to describe special texture images. With the development of image processing, 3D object description has been studied, and methods such as volume description, surface description, and generalized Cylinder description have been proposed. 6) image classification (recognition) belongs to the category of pattern recognition. The main content of the image is after some preprocessing (enhancement, restoration, compression, image segmentation and Feature Extraction for classification. Image classification uses classic pattern recognition methods, including statistical pattern classification and syntactic (structure) pattern classification, in recent years, the New Fuzzy Pattern Recognition and artificial neural network pattern classification have become increasingly important in image recognition.

Basic Features of Digital Image Processing
(1) Currently, most of the information processed by digital images is two-dimensional, with a large amount of processing information. For example, a 256x256 low-resolution black/white image requires about 64 kbit data size; for a high-resolution 512x512 image, the data size of kbit is required; if you want to process a TV Image Sequence of 30 frames per second, kbit ~ is required per second ~ 22.5mbit data volume. Therefore, the computing speed and storage capacity of computers are demanding. (2) digital image processing occupies a wide band. Compared with the language information, the occupied frequency band is several orders of magnitude larger. For example, the bandwidth of TV images is about 5.6 MHz, while that of voice images is about 4 kHz. Therefore, the implementation of imaging, transmission, storage, processing, and display is difficult and costly, which puts forward higher requirements for band compression technology. (3) The pixels in a digital image are independent and highly correlated. On the image screen, many pixels often have the same or close gray level. For a TV screen, the correlation coefficient of two adjacent pixels or two adjacent lines in the same line can reach more than 0.9, the correlation between two adjacent frames is larger than the intra-frame correlation. Therefore, there is a great potential for information compression in image processing. (4) because an image is a two-dimensional projection of a three-dimensional scene, an image itself cannot reproduce all the geometric information of a three-dimensional scene, obviously, some information behind a 3D scene cannot be reflected on a two-dimensional image. Therefore, to analyze and understand a 3D scene, you must make appropriate assumptions or add new measurements, such as binocular images or multi-viewpoint images. Knowledge guidance is required to understand 3D scenes. This is also a knowledge engineering problem that is being solved in AI. (5) After digital image processing, the images are generally observed and evaluated by people.
Factors have a greater impact. Due to the complexity of human visual systems, it is greatly affected by environmental conditions, visual performance, human hobbies, and knowledge. As a result, further research is needed to evaluate image quality. On the other hand, computer vision imitates human vision, and the perception mechanism of human will inevitably affect the research of computer vision. For example, what is the initial element of perception, how it is formed, the relationship between local and global perception, and the sensitive structure, attributes, and time features, these are the topics that psychology and neuropsychology are focusing on.

Advantages of Digital Image Processing
1. The fundamental difference between digital image processing and analog image processing is that digital image processing does not degrade image quality due to a series of transformation operations such as image storage, transmission, or copying. As long as the original image is displayed accurately during digitization, the digital image processing process can always reproduce the image. 2. High processing accuracy based on the current technology, it is almost possible to digitize a simulated image into a two-dimensional array of any size, depending on the capabilities of the image digitization equipment. Modern scanners can quantify the gray level of each pixel to 16 or higher, which means that the Digital Precision of the image can meet any application requirements. For a computer, no matter the size of the array or the number of digits in each pixel, its processing program is almost the same. In other words, in principle, no matter how high the image precision is, processing can always be achieved, as long as the array parameters in the program can be changed during processing. Looking back at the Analog Processing of a image, in order to increase the processing accuracy by an order of magnitude, it is necessary to greatly improve the processing device, which is extremely economic-cost-effective. 3. Applicable face width images can come from multiple sources of information, which can be visible images or invisible spectral images (such as X-ray images, ray images, ultrasonic images, or infrared images ). From the perspective of the objective object scale reflected by the image, it can be as small as electron microscope images, as large as aerial photos, remote sensing images, and even astronomical telescope images. After these images from different information sources are transformed into digital encoding formats, they are all gray-scale images represented by two-dimensional arrays (color images are also composed of gray-scale images, for example, RGB Images are composed of three gray images: Red, green, and blue. Therefore, they can be processed by computers. That is, as long as the corresponding image  measures are taken for different image information sources, the digital processing method of the image is suitable for any image. 4. Flexible and high image processing can be divided into three parts: image quality improvement, image analysis, and image reconstruction,
Each part contains rich content. In principle, image optical processing can only perform linear operations, which greatly limits the implementation of optical image processing. Digital image processing can not only complete linear operations, but also achieve non-linear processing, that is, all operations that can be expressed by mathematical formulas or logical relationships can be implemented by digital image processing.

Application of Digital Image Processing
Images are the main source of human acquisition and exchange of information. Therefore, the application of image processing will inevitably involve all aspects of human life and work. As the scope of human activity expands, the application of image processing will also expand. 1) the application of digital image processing technology in aerospace and aviation technologies, in addition to the JPL processing on the Moon and Mars photos described above, on the other hand, it is applied in aircraft remote sensing and satellite remote sensing technologies. Many countries send many reconnaissance planes to perform aerial photography in areas of interest on the Earth every day. We used to employ thousands of people to process and analyze the resulting photos. Now we use an image processing system with advanced computers to interpret the analysis, which saves both manpower and speed up the analysis, you can also extract a large amount of useful information that humans cannot find from photos. Since the end of 1960s, the United States and some international organizations have launched resource remote sensing satellites (such as LANDSAT series) and Sky Laboratories (such as Skylab ), the image quality is not very high because the imaging conditions are affected by the location, posture, and environmental conditions of the aircraft. Therefore, it is not cost-effective to obtain images at such an expensive cost. Instead, digital image processing technology must be used. For example, Landsat series Land satellites use multi-band scanners (MSS) to scan and imaging each area of the Earth at a height of 900km in a cycle of 18 days, the image resolution is roughly equivalent to a dozen or 100 meters on the ground (such as the LANDSAT-4 launched in 1983, the resolution is 30 m ). These images are first processed (digitalized, encoded) in the air into digital signals stored in the tape, when the satellite passes over the ground station, then high-speed transmission, and then analyzed and interpreted by the processing center. Many digital image processing methods must be used for these images, either during imaging, storage, transmission, or interpretation analysis. At present, all countries in the world are using images obtained by land satellites for resource surveys (such as forest surveys, marine sediment and fishery surveys, and water resource surveys ), disaster detection (such as pest and disease detection, water and fire detection, environmental pollution detection, etc.), Resource Survey (such as oil exploration, mineral volume detection, large-scale engineering geographic location exploration analysis, etc ), agricultural Planning (such as soil nutrition, water and crop growth and yield estimation) and Urban Planning (such as geological structure, water source and environmental analysis ). Our country has successively carried out some practical applications in the above aspects and achieved good results. Digital image processing technology has also played a significant role in meteorological forecasting and other space planet research.
2) the application of digital image processing in biomedical engineering is very extensive and effective. In addition to the CT technology introduced above, there is also a kind of processing and analysis of medical microscopic images, such as red blood cell, white blood cell classification, chromosome analysis, cancer cell identification, etc. In addition, image processing technology is widely used in medical diagnostics, such as X-ray lung image growth, ultrasonic image processing, ECG analysis, and three-dimensional radiation therapy. 3) communication engineering applications the main development direction of current communication is multimedia communication that combines sound, text, images and data. Specifically, the telephone, television, and computer are transmitted over the digital communication network in the form of three networks. Image Communication is the most complex and difficult, because the amount of image data is huge, for example, the speed of transmitting color TV signals is over 100 Mbit/s. To transmit such high-speed data in real time, encoding technology must be used to compress the bit volume of information. In a sense, encoding compression is the key to success or failure of these technologies. In addition to widely used entropy coding, DPCM Coding, and transformation coding, we are developing and researching new coding methods at home and abroad, such as branch encoding, Adaptive Network encoding, and wavelet transform image compression encoding. 4) Industrial and engineering applications in the industrial and engineering fields image processing technology has a wide range of applications, such as automatic assembly line inspection of the quality of parts, and Parts Classification, inspection of fault in printed circuit board, Stress Analysis of Elastic Mechanics photos, resistance and force analysis of fluid mechanics pictures, automatic sorting of postal mail, identification of the shape and arrangement of the workpiece and objects in some toxic and radioactive environments, and adoption of industrial vision in advanced design and manufacturing technologies. It is worth mentioning that the development of intelligent robots with visual, auditory, and tactile functions will bring new incentives to industrial and agricultural production, currently, it has been effectively used in spray paint, welding and assembly in industrial production. 5) military and public security applications in military aspect image processing and recognition are mainly used for precise terminal guidance of missiles and interpretation of various reconnaissance photos, automated military command systems with image transmission, storage, and display, simulated training systems for aircraft, tanks, and warships, interpretation and analysis of public security business images, fingerprint recognition, and face recognition, restoration of incomplete images, as well as traffic monitoring and accident analysis. Currently, the automatic identification of vehicles and license plates in the highway toll-free automatic charging system is an example of successful application of image processing technology. 6) Cultural and Artistic applications currently, such applications include digital editing of TV pictures, animation production, electronic image games, textile crafts design, clothing design and production, hair style design, the reproduction and restoration of cultural relics photos, athlete Action Analysis and scoring, and so on have gradually formed a new art-computer art.