Automatic Coloring Method For National Costume Sketches

The colorful ethnic costumes contain a rich national history and culture,and the colors with signs and symbols reflect the deeper national spirit and regional characteristics.Therefore,it is of positive significance to study the color laws and applications of ethnic costumes for the protection and inheritance of ethnic minority cultures.At present,there are some problems in this research areas,such as the low degree of technology applications and reliance on professional experience.It severely restricts the development and utilization of ethnic cultural resources.This makes to use new computer technology to explore and study automatic image coloring methods for the application of digital resources for ethnic apparel has certain research value and application prospects.In this paper,we construct a diverse costume sketch database and use the advantages of generative adversarial networks in image generation.The neural network is trained to get a generator for coloring sketches.In this way,the automatic coloring method for hand-drawn sketches is realized.The main research contents are as follows:(1)An algorithm for extracting edges of national costume patterns was designed.This paper proposes a new edge detection algorithm based on discrete H-channels.This algorithm defines multi-scale color labels in HSV space,converts the original image into H-channel color label images.Discrete similarity of area is calculated to determine edge points.The entire detection algorithm process is completed.Experimental analysis shows that the algorithm can greatly improve the detection accuracy of traditional gradient method.(2)A national costume sketch library was built.This paper uses the proposed edge detection algorithm to process the national costume image.Based on the obtained national costume edge map,it performs multiple styles of sketching processing by the existing sketch generation algorithm.It generates a database by filtering.This database of national costume images includes color images and hand-drawn sketches.It provides important data basis for the research of digital costume related fields.(3)A coloring model of national costume sketch was constructed.This paper uses the advantages of deep learning technology to build a national costume sketch coloring model based on generative adversarial networks.This model consists of a 6-layer Unet structure,and the discriminator consists of a 5-layer convolutional neural network.Based on this,the loss function and the weights of true reliability value in the discriminator are optimized.The data of the sketch database is used for model training to get an automatic coloring model.The results of a large number of sketch coloring experiments show that the model has a good learning ability for the color law of national costumes and can achieve better clothing coloring effects.(4)An automatic coloring prototype system for national costume sketches was implemented.This paper designs and implements a prototype system for displaying the coloring effect of national costumes sketches.This system integrates functions such as edge detection,sketch generation and sketch coloring.It visualized the staged results of this research and improves the effectiveness and practicability of proposed method.