Overview of Digital Image Processing
Digital Image Processing (Digital Image Processing), also known as computer image processing, refers to 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. image processing methods that are frequently used 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 aerospace Inspector 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, more complex image processing was performed on the nearly 100,000 photos sent back by the exploration ship. As a result, the topographic map, color map, and panoramic mosaic map of the moon were achieved, with 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 exploration of Mars, Saturn and other planets. Another major achievement of digital image processing is the achievement of medicine. In 1972, housfield, a project engineer at the British EMI company, invented the X-ray computer tomography device for skull diagnosis, which we generally call CT (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 projects, industrial inspection sites, 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 beginning 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 similar human visual systems understand the external world. This is called image understanding or computer vision. Many countries, especially developed countries, have invested a lot of other manpower and material resources in this study, and have achieved many important research results. 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 area 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 of digital image processing is digital image processing. The main research content includes the following aspects: 1) image transformation because the image array is very large, it is processed directly in the spatial domain, it involves a large amount of computing. Therefore, NLP often uses various image transformation methods, such as Fourier transform, Fourier transform, discrete cosine transform, and other indirect processing technologies to convert the processing of spatial domains into transform domains, this not only reduces the computing workload, 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 volume of data describing the image (that is, the number of BITs), so as to save image transmission, processing time and reduce the memory capacity occupied. Compression can be obtained without losing sight of the truth, and can also be performed under the distortion of consent. Encoding is the most important method in compression technology. It is the earliest and more 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 use a certain filtering method, restore or recreate the original image. 4) image Cutting Image cutting is one of the key technologies in digital image processing. Image cutting 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 area cutting methods have been developed, there is no effective method that is applicable to various images. Therefore, the research on image cutting is still in depth, which is one of the hot topics in eye image processing. 5) image description Description Image Description description is a prerequisite for image recognition and understanding. As the simplest binary image, you can use its geometric characteristics to describe the characteristics of an object. Generally, the descriptive method of an image is described by two-dimensional shape, it has two methods: Border description and regional description. For special texture images, two-dimensional texture features can be used to describe them. With the development of image processing research, we have begun to study the description of 3D objects, and put forward methods such as volume description, surface description, and generalized Cylinder description. 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 cutting and Feature Extraction for classification. Image classification often 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) At present, most of the information processed by digital images is two-dimensional, and the amount of information processed is very large. 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; for example, 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, there are often many pixels with 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 3D scenes, you must make appropriate assumptions or add new attention volumes, such as binocular images or multi-viewpoint images. Knowledge guidance is required to understand 3D scenes. This is also a knowledge project that is being resolved in AI. (5) After digital image processing, images are usually observed and evaluated. Therefore, they are greatly influenced by human factors. Because human visual systems are very complex and greatly influenced by environmental conditions, visual performance, human hobbies, and knowledge conditions, further research is needed to evaluate image quality. On the other hand, computer vision is a visual imitation of people, and the perception mechanism of people 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 simulated 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. Only when the image is digitalized to accurately display the original image, the digital image processing process can always maintain the reproduction of the image. 2. The processing accuracy is high based on the current technology. It is almost possible to digitize a simulated image into a two-dimensional array of random size, depending on the capabilities of the image digitization device. 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, the processing can always be achieved. You only need to change the number of arrays in the program during processing. In order to increase the processing precision by an order of magnitude, We need to greatly improve the processing device, which is extremely economic. 3. Wide face images can come from multiple sources of information, which can be visible or invisible spectral images (such as X-Ray, Ray, ultrasonic or infrared images ). From the objective object scale reflected by the image, it can be small to electron microscope images, large to 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 to say, we only need to extract the corresponding image information collection measures for different image information sources. The digital processing method of images is applicable to any image. 4. Flexible and high image processing can be divided into three parts: image quality improvement, image analysis, and image reconstruction. Each part includes 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.
The application of digital image processing is the main source of human acquisition and exchange of information. Therefore, the application of image processing must 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 a large number of reconnaissance planes to perform aerial photography in areas of interest on the Earth every day. We have to employ thousands of people to process and analyze the resulting photos. Now we use an image processing system with advanced computers to interpret and analyze images, which saves both manpower and speed up the process, it can also extract a large amount of practical information that cannot be found by humans 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 ), because the imaging conditions are affected by the location, posture, and environmental conditions of the aircraft, the image quality is not always very high. 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, 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 quickly transmitted, and then analyzed and interpreted by the processing center. Many digital image processing methods must be applied to images, storage, transmission, and interpretation analysis. Today, 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 indexes (such as pest and disease detection, water and fire detection, and environmental pollution detection ), resource Surveys (such as oil exploration, mineral exploration, and geographic location Exploration and Analysis of large-scale projects), Agricultural Planning (such as soil nutrition, water and crop growth, and yield estimation ), city 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 research on other planets in space. 2) the application of digital image processing in the biomedical project is very extensive and very effective. In addition to the CT technology described above, the other is the processing and analysis of medical microscopic images, such as red blood cell, white blood cell classification, chromosome analysis, and cancer cell identification. 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) applications of communication projects currently the main development direction of communication is multimedia communication that combines sound, text, images and data. In detail, 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, such as the speed of transmitting color TV signals reaching 100 Mbit/s or more. To transmit such fast data in real time, the bytes must use encoding technology 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 encoding, and transform encoding, we are developing and researching new encoding methods at home and abroad, such as branch encoding, Adaptive Network encoding, and wavelet transform image compression encoding. 4) Industrial and project applications in the industrial and project fields image processing technology has a wide range of applications, such as self-active 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, self-sorting of postal mail, identification of the shape and arrangement of the workpiece and objects in some toxic and radioactive environments, and the use 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, spray paint, welding, and Assembly have been effectively used 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, automatic Military Command System with image transmission, storage and display, simulated training system 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, active identification of vehicles and license plates in the Self-charging system for no parking on the expressway is an example of successful application of image processing technology. 6) applications of culture and art at present, 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.
Brief Introduction to digital image processing
