Image Engineering, Volume 1: Image Processing and Analysis

Author: Zhang Yujin
Editor-in-Chief: Wang Renkang
Publisher:
Publishing Date: 2005-05-01
Features:
Content Summary
This book primarily introduces the fundamental principles, typical methods, and practical techniques of image processing and analysis. Considering the rapid development and widespread application of image technology, the book not only explains basic theories but also presents many recent new research achievements and application examples from around the world. The book is divided into three main parts.
Part 1 (Chapters 1, 2, and 3) covers image fundamentals, discussing the definition of image engineering, an overview and classification of image technology, as well as visual and image models. It also covers digital image acquisition, representation, pixel relationships, and various basic transformation techniques.
Part 2 (Chapters 4, 5, and 6) explores important branches of image processing, such as image enhancement, image restoration, projection-based image reconstruction, and image compression coding, along with their fundamental theories, techniques, and methods.
Part 3 (Chapters 7, 8, and Appendix A) introduces the fundamental principles and techniques of image analysis, including image segmentation, object representation and description, feature measurement, and morphological methods. The book also provides numerous examples and exercises. This book can serve as a textbook for undergraduate and graduate students in disciplines such as information and signal processing, communications and electronic systems, pattern recognition, and biomedical engineering. It is also suitable for researchers and educators in the aforementioned fields, as well as in information engineering, electronic engineering, computer science and technology, robotics and automation, remote sensing, and military reconnaissance.
Excerpt:
A basic image (processing and analysis) system can be composed of the modules shown in Figure 1.2.1. Each module has a specific function, including acquisition, display, storage, communication, processing, and analysis. The figure also lists some equipment required to perform these functions. Image acquisition can utilize charge-coupled devices (CCDs), video cameras equipped with vidicons, and scanners, among others. Image display can be achieved using television monitors (TV monitors), cathode-ray tubes (CRTs), and various printers. Image storage can involve magnetic tape, magnetic disks, optical disks, and magneto-optical disks. Image communication can rely on integrated services digital networks (ISDN), computer local area networks (LANs), or even conventional telephone networks (PSTNs). Finally, image processing and analysis primarily involve computations, with computers being the main equipment used, though specialized hardware may be employed when necessary.
1.2.2 Image Acquisition Module
To acquire digital images, two devices are required. One is a physical device sensitive to a specific electromagnetic energy spectrum band (e.g., X-rays, ultraviolet, visible light, infrared), which generates an analog electrical signal proportional to the received electromagnetic energy. The other is a digitizer, which converts the aforementioned analog signal into a digital (discrete) form. All equipment for acquiring digital images requires both of these devices.
Taking the common X-ray imaging system as an example, X-rays emitted from an X-ray source pass through an object and reach a media sensitive to X-rays on the other side. This media captures an image of the object material's varying absorption of X-rays. It can be film, a television camera that converts X-rays into photons, or other discrete detectors capable of outputting digital images. Equipment for visible light and infrared imaging mainly includes microdensitometers, image dissectors, vidicons, and photon-sensitive solid-state arrays.
When using a microdensitometer, the image to be digitized should be in the form of a transparent slide or photograph. Vidicons and photon-sensitive solid-state arrays, in addition to accepting these forms of images, can also digitize natural images with sufficient light intensity that enter the detector. When using a microdensitometer, the transparent slide or photograph needs to be placed on a flat plate or wound on a cylindrical drum. As light focuses on the image, moving the plate or rotating the drum completes the scanning. For a transparent slide, light passes through it; for a photograph, light reflects off its surface. In both cases, the light beam focuses on a photon detector, and each detector records the image's grayscale value corresponding to its current position. If both the grayscale value and position coordinates are taken as integers, a digital image is obtained. Although microdensitometers are relatively slow, their spatial accuracy is high due to the continuous nature of the mechanical translation process.
The working principle of a vidicon camera is based on photoconductive properties. An image focused on the surface of a vidicon forms a photoconductive pattern corresponding to the brightness distribution of the optical image. Another independent electron beam scans the other side of the photoconductive tube, and due to neutralization, this beam generates a voltage difference signal on a receiver that corresponds to the input light intensity pattern. If this signal is quantized and the corresponding scan beam position is recorded, a digital image is obtained.
A solid-state array consists of discrete silicon imaging elements called photosites. Such photosites generate an output voltage proportional to the input light intensity received. Solid-state arrays can be categorized into two types based on their geometric organization: line scanners and plane scanners. A line scanning sensor includes a row of photosites and acquires a 2-D image through the relative motion between the scene and the detector. A plane scanning sensor, composed of photosites arranged in a matrix, directly obtains a 2-D image. A significant feature of solid-state plane sensor arrays is their very fast shutter speed (up to 10?? s), allowing them to freeze many movements.
The main component of a solid-state array is the charge-coupled device (CCD). Figure 1.2.2 shows a schematic diagram of a line-scanning CCD sensor. This sensor consists of a row of photosites, two transfer gates that periodically transfer the content of the photosites to a transfer register, and an output gate that periodically transfers the content of the transfer register to an amplifier. The voltage signal output by the amplifier is proportional to the content of the photosite row. The working principle of a charge-coupled plane array is similar to that of a line array, but here the photosites are arranged in a matrix form (as shown in Figure 1.2.3). The columns of photosites are separated by transfer gates and transfer registers. First, the content of the odd-numbered photosite columns is sequentially sent into the vertical transfer register and then into the horizontal transfer register. The content of the horizontal transfer register is sent to the amplifier to obtain a interlaced video signal for one frame. Repeating the same process for the even-numbered photosite columns yields another interlaced video signal for another frame. Combining the two frames results in one field (f) of interlaced scanning television. The scanning speed for the NTSC system is 30 f/s, while for the PAL system, it is 25 f/s.
Currently, line-scanning CCDs commonly have 512 to 4,096 pixels or more, and plane-scanning CCDs with 4,096 × 4,096 pixels are also in use. Using image processing and analysis techniques, larger field-of-view images can be obtained with lower-resolution CCDs through image stitching [Zhang 1997b]. Television cameras are typically composed of CCD arrays, and to obtain a digital image, the camera's video output must be sent to a digitizer. This is often achieved by inserting a specialized hardware card into a computer.