Study On Plant Simulation And Image Compression Techniques Based On Fractal

Fractal geometry, which is one of the growing points in non-linear science, focuses on describing objects with irregular shapes, and is characterized by self-similarity in external structure and simplicity in internal essence. Fractal geometry is one of the best mathematical tools for people to describe and simulate various beautiful natural scenes, thus being known as the natural geometry. Much research has been conducted on plant simulation based on iteration function system (IFS) and L system. However these methods still need to be developed and improved, for now it is difficult to simulate various plants with different shapes realistically and naturally, moreover the forming mechanism and rules of the natural scenes also need to be further studied. On the other hand, the inversion of IFS has been applied to image compression, which is the key technique for multi media communication by Internet. However the traditional PIFS image compression method suffers from the long time consuming problem and the insufficient utilization of fractal characteristics of image, which limits the fractal based compressing methods for practical applications. Aiming at the two above-mentioned problems, this paper investigates the realistic plant simulation methods and the rapid fractal based image compression techniques. The following research achievements have been accomplished:(1) Tree Simulation Based on IFSThe existing IFS based plant simulation approaches mainly concentrates on generating natural plants in monochrome image form. Consequently there exist no reasonable rendering methods for color pants simulation. Moreover, the simulated plants by IFS based method are lack of variation due to the insufficient understanding of topological structure of plant. To solve the above problems, this paper explores several IFS based simulating techniques for color tree. One is the natural rendering method for a color tree. And the other is the tree shadow drawing method in natural illumination environment. Furthermore, in order to simulating the different and variant spatial structure of tree, the influences of iteration probability and stochastic factors on tree spatial structure and configuration are studied. Finally, The simulating experiments are conducted for establishing the feasibility and effectiveness of the proposed approaches.(2) Plant simulation based on fusion of L system and IFSThe general fractal based methods for plant simulation are L system and iteration function system (IFS). It is difficult to realize the natural and realistic plant simulation only by use of the traditional L system or iteration function system (IFS) separately owing to the limitation of each method. In this paper, the main characteristics of both methods are analyzed thoroughly, and a novel plant simulation method based on fusion of L system and IFS is proposed, so that the merits of different approaches can be combined. Moreover, an improvement is made on the traditional L system algorithm for the convenient implementation of the proposed fusing method. The simulating results demonstrate that the presented method can simulate plantsmuch more freely, naturally and realistically.(3) Study on fast fractal image compression techniquesThrough analyzing the characteristics of the traditional PIFS coding method and its corresponding improvement approaches, this paper present a fast fractal image compression method based on block classification to overcome the long time consuming problem of the existing compression methods. In this method, the means and standard derivations of energy of each block in horizontal, vertical and diagonal DCT frequency domain respectively are employed for the classification features. The WeiSC method and the Kohonen neural network are used to classify image blocks. The experimental results show that the proposed method can classify blocks more accurately than the traditional PIFS coding method, and the coding speed of this method has increased near 10 times compared with the conventional global search method.