Research On The Image Analysis Algorithms Based On Shape Feature

In this thesis I gave the development level and the new trend in digital image processing and retrieval national and international. This work was done based on lots of the relevant references in this field. The author studied the main techniques for image retrieval and analyzed the shape feature extraction and description method. Furthermore, in this thesis I discuss in details about the shape feature based image analysis algorithm and realized it in the experiment. Discussion and analysis on these kinds of algorithm were given in the thesis. The thesis is organized in five sections as follows.The first section is the Introduction. In this section, analysis and discussion about the development of the history, current status and the developing trend in image processing from national and international view are given. In the main part I introduce the characteristic, the superiority, the main utilization and the main algorithm, the technique and the current developing status, main utilization. Theory and the realization technique for shape based image retrieval are given.The second section is shape based image feature description. In an image, the characteristic of the object should not be changed with the transposition, rotation and scale. For this reason, when to describe a shape, the selected features should sustain these kinds of change and keep invariant. Besides this, the selected descriptor should be able to describe the essence of the shape and have good discrimination. There are many methods for shape analysis and classification. In general, shape features can be divided into two categories, boundary based and region based. The classical methods for shape descriptions are Fourier descriptor, shape variant matching module and shape invariant moment. In this section, I study the classical invariant moment, boundary invariant moment and affine invariant moment. Shape invariant moment is a kind of linear feature, which is invariant for image rotation, scale and transposition. Particular discussion on affine invariant moment was given. From above, we can conclude that the image region can be described through shape invariant moment.