Application Research On Image Editing And Propagation Based On Color Diffusion In The Mechanical Design

With the popularity of people-oriented design concept,mechanical design more and more reflects the artistry and individuation of products.The development of image editing and propagation technology has saved a lot of time cost and manpower cost for mechanical design.In this paper,an image colorization method based on the region growing theory and graph theory is proposed.Besides that,an image editing and propagation method based on self-adaptive mutation particle swarm optimization and support vector machine is proposed.The main research results are as follows:1.In this paper,an image colorization method based on the region growing method and graph theory is proposed.Firstly,the foreground features of the target image are extracted by the region growing method.And then,in the extracted target image,the similarity weights matrix on the structure relationship between the pixels is obtained according to graph theory.Finally,color diffusion is realized by the similar weight matrix.The traditional region growing method often causes inaccurate feature extraction due to improper seed points.In this paper,the principle of maximum similarity is used to solve it.That can improve the accuracy of foreground feature extraction,and reduce the misstaining rate of the target region.Thus,it can make the colorization more accurate.2.An image editing and propagation method based on self-adaptive mutation particle swarm and support vector machine(SVM)is proposed.Firstly,editing propagation is transformed into a classification problem of SVM in feature space.The colorization can be realized under SVM.Then,the self-adaptive mutation particle swarm optimization algorithm is used to optimize SVM,which avoids the problem of large data classification error caused by the improper parameter of SVM.In the color diffusion,reconstructing the local relationship between the pixels according to the graph theory to make colorization more natural.And then distinguishing the contribution of different pixels by the similarity weights among the pixels.So the colorization will be more natural,more realistic.The experimental results show that the proposed methods can obviously improve the speed of the algorithm while keeping the high fidelity staining of the image.It has a good effect in grayscale image colorization,color image secondary dyeing and so on.