计算机视觉+BookList.pdfVIP

12 CHAPTER 1 INTRODUCTION To illustrate the systematic procedures of an image processing system, we give an example of human face identification (Adler and Schuckers, 2007). The problem domain is the faces of people. The objective is to associate the face with the identity of the person. The output is a person’s unique identifier (e.g., social security number). The necessary procedures to achieve the goal of face identification could include 1. Image acquisition: The face image could be acquired through a high-resolution still digital camera and compressed to an image file. 2. Preprocessing: The acquired image may be enhanced by improving contrast, sharpness, color, and so on. 3. Segmentation: The image may first be cropped to only the facial area. Then, the face may be segmented into eyes, mouth, nose, chin, and so on. 4. Representation and description: In this step, each of the segmented areas may be characterized by statistical data, for example, principal components ana- lysis, texture, aspect ratios of eyes and nose, or the color of eyes. 5. Matching recognition and interpretation: This step may involve using the characteristics derived in the previous step to match each individually seg- mented area based on specific recognition algorithms. For example, eyes may be processed to determine, based on its features, what class of eye it is. Then, all of these interpretations are used to create a composite description of the “ensemble,” perhaps in the form of a feature vector for the subject. 6. Knowledge base: Finally, the feature vector above may be fed to a knowledge base of all known subjects to associate it with one of the subjects in the database, thus returning perhaps the individual’s social security number or perhaps a confidence score of the match. APPENDIX 1.A SELECTED LIST OF B