Analysis of Visual Expression of Light and Color in Movies based on Wavelet Neural Network

Movies are in the era of “black, white and gray” ruling the screen, people have a strong desire for movies to truly reproduce colors, because the main rhythm of the “black, white and gray” film visual music is not clear enough, the momentum is not strong enough, the meaning is not rich enough, and the color is not enough. Visual aesthetics cannot be attractive… Aiming at the dual challenges of colorization quality and timing stability in film colorization, a generative adversarial network with a recurrent structure is proposed, which can be used for automatic colorization of films without any reference frames and artificial intervention. The network generates a confrontation network based on classical conditions: the generator is used to generate color images and complete the task of colorization; The discriminator is used to identify the authenticity and improve the performance of the generator. Cycle structure and timing consistency loss are introduced to integrate timing information and solve the stability problem of coloring. Experimental results show that this method can effectively reduce flicker in the generated movie sequence while ensuring the coloring of a single frame image. This paper studies and analyzes the visual expression of light and color in movies based on wavelet neural network.