Image enhancement is the processing of images to improve their appearance to human viewers or to enhance other image processing systems’ performance. Methods and objectives vary with the application. When images are enhanced for human viewers, as in television, the objective may be to improve perceptual aspects: image quality, intelligibility, or visual appearance. In other applications, such as object identification by machine, an image may be preprocessed to aid machine performance. Because the objective of image enhancement is dependent on the application context, and the criteria for enhancement are often subjective or too complex to be easily converted to useful objective measures, image enhancement algorithms tend to be simple, qualitative, and ad hoc. In addition, in any given application, an image enhancement algorithm that performs well for one class of images may not perform as well for other classes.
Image enhancement is closely related to image restoration, which will be discussed in Chapter 4. When an image is degraded, restoration of the original image often results in enhancement. There are, however, some important differences between restoration and enhancement,
· in image restoration, an ideal image has been degraded, and the objective is to make the processed image resemble the original image as much as possible.
· In image enhancement, the objective is to make the processed image better in some sense than the unprocessed image.
In this case the ideal image depends on the problem context and often is not well defined. To illustrate this difference, note that an original, undegraded image cannot be further restored but can be enhanced by increasing sharpness through highpass filtering.
Image enhancement is desirable in a number of contexts. In one important class of problems, an image is enhanced by modifying its contrast and/or dynamic range. In another class of enhancement problems, a degraded image may be enhanced by reducing the degradation. Another important class of image enhancement problems is the display of 2-D data that may or may not represent the intensities of an actual image.
1.An image is enhanced by modifying its contrast and/or dynamic range. For example, a typical image, even if undegraded, will often appear better when its edges are sharpened. Also, if an image with a large dynamic range is recorded on a medium with a small dynamic range, such as film or paper, the contrast and therefore the details of the image are reduced, particularly in the very bright and dark regions. Contrast in an image taken from an airplane is reduced when the scenery is covered by cloud or mist, increasing the local contrast and reducing the overall dynamic range can significantly enhance the quality of such an image.
2.A degraded image may be enhanced by reducing the degradation. Examples of image degradation are blurring, random background noise, speckle noise, and quantization noise. This area of image enhancement overlaps with image restoration.
An algorithm that is simple and ad hoc, and does not attempt to exploit the characteristics of the signal and degradation, is generally considered an enhancement algorithm. An algorithm that is more mathematical and complex, and exploits the characteristics of the signal and degradation with an explicit error criterion that attempts to compare the processed image with the original undegraded image, is generally regarded as a restoration algorithm. This distinction is admittedly somewhat vague and arbitrary. Some arbitrary decisions have been necessary in dividing certain topics between this chapter and the next chapter, which deals with the image restoration problem.
It is well known that the contours or edges in an object contain very important information that may be used in image understanding applications. The first step in such an application may be to preprocess an image into an edge map that consists of only edges. Since more accurate detection of edges in an image can enhance the performance of an image understanding system that exploits such information, converting an image to its corresponding edge map may be viewed as an enhancement process.
3.Another important class of image enhancement problems is the display of 2-D data that may or may not represent the intensities of an actual image. A low-resolution image of 128 x 128 pixels may be made more visually pleasant to a human observer by interpolating it to generate a larger image, say 256 x 256 pixels. In 2-D spectral estimation, the spectral estimates have traditionally been displayed as contour plots. Although such 2-D data are not images in the conventional sense, they can be presented as images. We can display them as black-and-white images, or we can enhance them with color so that their appearance may be improved and information conveyed more clearly. In other applications, such as infrared radar imaging, range information as well as image intensities may be available. By displaying the range information with color, relative distances of objects in an image can be highlighted. Even good-quality images may be enhanced by certain types of distortion. For example, when an object in an image is displayed with false color, the object may stand out more clearly to a human viewer.