Research On Image Fusion Technology For Customer Experience

In order to reduce the fusion trace of the face images and improve the fusion effect,an expression fusion method based on reference face and an improved adaptive nonlinear weighted fusion algorithm based on feature point detection are proposed.In virtual shopping,in order to make customers see the effect of jewelry wearing intuitively,a virtual jewelry trying method is introduced.In order to improve the shopping experience of the customer,a face gradient special effect is added during the face image replacement process.In view of the problem that the commonly used face feature point detection methods lack of forehead feature point detection,the convolution neural network is used to realize the forehead feature point detection,which effectively compensates for the lack of detecting the forehead feature information by only detecting the internal feature points on the face.In view of the problem that traditional image fusion methods lack an adaptive mechanism,when the face image is fused,according to the brightness difference between the fusion images,the weighting functions of the fusion region can be adaptively adjusted to smooth the change speed of the fusion boundary and reduce the fusion trace.By merging the adjusted customer face image and model image with different weights,a gradient special effect is added when the face image is replaced.After the face image replacement is completed,the jewelry image is spliced onto the model by threshold segmentation and perspective transformation methods to try on the jewelry.The experimental results show that convolution neural network can effectively extract the forehead feature points from the fusion images,and the proposed adaptive nonlinear weighted fusion can effectively improve the fusion effect of the face image.The face alignment results based on 70 feature points are better than that based on68 feature points which are extract by Dlib library.The stitching traces on the forehead area could be improved.The adaptive nonlinear weighted fusion method can further reduce the fusion traces,and the gradients in different directions of the image fusion region can be reduced,and the natural transition of image fusion can be obtained.The expression fusion method based on the reference face can make the expression richer in the result of face replacement.Through virtual jewelry wearing,customers can intuitively view the effect of wearing jewelry.