Texture Based Image Segmentation and analysis of medical image

Dealing with information extracted from a natural image, a medical scan, satellite data or a frame in a video sequence is the purpose of image analysis. In the real world, the stimulus that is received by the retina is perceived as whole and complete information. Between the electromagnetic reception and the perception, physiological and neurological processes construct the final perception and analysis of the image. In fact vision is composed of many interacting components including analysis of color, texture and shape, the whole conducted by prior knowledge of the human brain. Computer vision aims at getting the same result as human perception. The computer interface receives the image as a matrix of pixels/voxels and several levels of processes are involved to get, when it is possible, the same result as human analysis. The collection of processes involved in the visual perception are usually hierarchically classified as belonging to either low level vision or high level vision. High level vision consists of the interpretation of the image following some rule or prior knowledge. In low level vision, image processing is performed to extract some visible physical properties in the image such as shape and boundaries or to improve the quality of the image. In this thesis we will be dealing with image processing and more precisely with the image segmentation task. The objective of segmentation methods is to determine a partition of an image into a finite number of semantically important regions such as anatomical or functional structures in medical images or objects in natural images. The segmentation task has been studied for several decades; however it is still a challenging task. This task is essential in many applications including face detection in video sequences, changes detection in satellite images, anatomical or functional object extraction in medical images or object extraction in natural images.