In view of the loss of color information,disordered stroke details,an d distortion of edge lines in the traditional image stylization processing algorithm,this paper improves the image style transfer algorithm based on convolutional neural network,and comprehensiv e consideration of image brightness and color information characteristics.This paper proposes a n image style transfer algorithm based on a new style loss function and brightness information .First,the content image is converted into the color space,and the style transfer is performed on the brightness channel alone,retaining the color information of the content image.Second,the histogram loss is added to the total loss function to construct a new total loss function,which i mproves the stability of the stylized image.Final,the edge information of the extracted content imag e is superimposed on the result image to achieve the purpose of improving the vagueness of the edg e information of the stylized result image.Experiments show that the improved method has better overall performance in stroke details,texture structure and color space.