Image Enhancement Techniques: Zooming and Super Resolution

Today, after more than one hundred year of perfecting photographics techniques, the majority of visual information received by a human are of a very high quality. This requests for any new imaging technology to generate images of comparable quality, in other words images that have a very high resolution. The resolution of an image is dependent on the resolution of the image acquisition device, i.e. the number of photo-detectors. Sensors with high density of photo-detectors capture image at a high spatial resolution. Sensors with fewer photo-detectors produce a lower resolution images leading to pixelization where individual pixels are discernable at naked eye.
One way to increase the sampling rate is to increase the number of photo-detectors and to decrease their size thereby increasing the sensor density. There is unfortunately a limit beyond which the shot noise degrades image quality irremediably. It is hence necessary an algorithmic way to enhance the resolution of the camera once the resolution of an image sensor has been pushed to a reasonable economical and technological limit.
The immediate solution is to resize an image and approximate by interpolation the missing samples. The first part of this thesis is hence concerned with improvements over the standard interpolation algorithms for images. In particular several innovative zooming techniques are introduced.
The quality of the interpolated image generated by a single image in any case is limited by the amount of data available. Image zooming cannot produce high frequency components lost during the low-resolution sampling process. Pragmatic improvements may be obtained starting with multi-input data sets in which additional data constraints from various observations of the same scene allows a super-resolved reconstruction of the scene. Each low-resolution observation from neighboring frames potentially contains novel information about the target high-resolution image. In the second part of the thesis are reported different super resolution techniques based on multi input images.
The rest of the thesis is devoted to report other image enhancing, image coding, low level image processing algorithms that have been produced during my graduate student years. In detail the rest of the thesis is organized as follows: in chapter 1 is reported a description of image acquisition devices, with particular interest on the digital still camera. Chapter 2 reports different zooming techniques, and a comparison between these methods. In chapter 3 are reported some super resolution techniques and, in particular, the description of a new approach to merge different frames. Chapter 4 reports a re-indexing algorithm to color mapped images, and a method to find the edges in noisy picture. Concludes the chapter the description of a new technique for automatic discrimination of text images.