image segmentation

Category: Education

Presentation Description

No description available.


By: fkl_shohag (38 month(s) ago)

Extra ordinary work...thks.

Presentation Transcript



Image Segmentation : 

Image Segmentation Image segmentation refers to the process of partitioning a digital image into multiple segments .It is typically used to locate objects and boundaries in images. More precisely, it is the process of assigning a label to every pixel. The result of image segmentation is a set of segments that collectively cover the entire image, or a set of contours extracted from the image. Each of the pixels in a region are similar with respect to some characteristic or computed property, such as color, intensity, or texture. Adjacent regions are significantly different with respect to the same characteristics Image segmentation

Medical Image Segmentation : 

Medical Image Segmentation The Image segmentation is one of the most important parts of clinical diagnostic tools. Medical image segmentation refers to the segmentation of known anatomic structures from medical images. Structures of interest include, tumors and cysts, as well as other structures such as bones, vessels, brain structures . The overall objective of such methods is referred to as computer-aided diagnosis; in other words, they are used for assisting doctors in evaluating medical imagery or in recognizing abnormal findings in a medical image. Segmentation results of brain

Methods of Segmentation : 

Methods of Segmentation Images are often interfered by signals and artifacts which rose of during sampling, what may cause big problems at using of common techniques of segmentation. Therefore, accurate segmentation of medical images is a very difficult task. However, the process of accurate segmentation of these images is very important and crucial for a correct diagnosis by clinical tools. Many methods have been developed for better image segmentation. In this paper we are going to make a comparative study of three methods namely Active Contour Method, Bayesian method and Biomimetic Pattern Recognition Method.

Active Contour Method : 

Active Contour Method Besides challenges due to imaging noise and partial volume effects, the similarity in intensity and texture between neighboring structures complicates the task of identifying distinct boundaries between the structures. So the active contour method was introduced which developed the concept of shape contours .When evolving shape contours, the interaction consists of modeling the “forces” of attraction, repulsion, and competition by taking into account the relationship between object contours and their shape estimates.

Modes of Interaction : 

Modes of Interaction In Active Contour, segmentation is performed by iteratively repeating two interactive operations: Contour evolution:When evolving shape contours, the interaction consists of modeling the “forces” of attraction, repulsion, and competition by taking into account the relationship between object contours and their shape estimates. The attraction describes the force of drawing the organ contour toward the learned shape prior, while repulsion and competition define the actions between neighboring curves to avoid overlapping and to solve the ambiguity of which structures the voxels belong to.

Slide 7: 

Posteriori shape estimation method :The shape priors are generated according to shape prior distribution, neighboring shapes, image features, and also the current evolved curves. Energy functionals are then formulated to model the interactions. Segmentation is achieved by minimizing these functionals.With the proposed approach, neighboring structures with similar intensities and/or textures, and blurred boundaries can be extracted simultaneously.

Bayesian Method : 

Bayesian Method The Bayesian method provides a way to solve image Reconstruction problems that would otherwise be insoluble . The Bayesian approach is based on probability theory, which makes it possible to rank a continuum of possibilities on the basis of their relative likelihood or preference and to conduct inference in a logically consistent way. Segmentation by bayesian method

Naïve Bayes Model : 

Naïve Bayes Model The naïve conditional independence assumption allows efficient computation of marginal and conditional distributions for large-scale learning and inference. We choose a generative model over the discriminative counterpart motivated in part by a faster convergence rate of the asymptotic generalization error when label information is scarce. The naive Bayes classifier finds successful application in text categorization tasks . The Naive conditional independence assumption allows us to factorize the joint distribution as a product of class prior and independent conditional probability

Biomimetic Pattern Recognition Method : 

Biomimetic Pattern Recognition Method Based on mathematic topological analysis of the sample set in the high dimensional feature space, this model utilizes the continuity of the same class of samples in the feature space. In the BPR theory, the construction of the subspace of a certain type of samples depends on analyzing the relations between the trained types of samples and practicing the methods of “coverage of objects with complicated geometrical forms in the multidimensional space”. The BPR is based on “pattern cognition” instead of “pattern classification” The method firstly uses neuron networks to completely cover the samples‟ high dimensional feature space and then segment medical images based on the results of the optimal coverage of the samples.

Multi-weight neuron network architecture : 

Multi-weight neuron network architecture Here, BPR is realized by Multi-weight Neuron Networks. In training of a certain class of samples, a Multi-weight Neuron Sub-network should be established. The sub-network has three layers: Layer 1 (input layer): this layer is used for system data input, which means sample feature vector. Layer 2 (Multi-weight Neuron hidden layer): this layer uses closed hyper surface to cover input sample feature vector in the high dimension. The number of Multi-weight Neurons in this layer of sub-network is determined by the number of training samples. Layer 3 (output layer): this layer is composed by i nodes, in this paper i =3, which represent CFS, GM and WM respectively. If the current input sample belongs to i-class,

Conclusion : 

Conclusion Here the Active Contour Method is a novel method that is able to effectively segment neighboring structures with similar texture and intensity. By intuitively describing the relationships between neighboring structures and their respective shape estimates our method effectively models interaction between neighboring contours to enable extraction of the boundaries that separate them. In Baysens method by imposing the transductive learning and inference problem w.r.t. time and in conjunction with spatiotemporal regularization constraints efficient segmentation data could be achieved. Multimodal registration was applied to bring the multimodal data sources into a common coordinate frame. In the Biomimetic pattern, each class of samples is trained to be “cognized” one by one, which produces better segmentation. By the comparative study of all three methods we have concluded that each method have its own advantages and disadvantages and so each method can be used in image segmentation according to the area of interest.