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

In the current era of film, characterised by a predominance of black, white, and grey tones, there exists a notable yearning among individuals for films to faithfully replicate colours. This desire stems from the perceived shortcomings of films predominantly composed of black, white, and grey visuals, which are often criticised for their lack of clarity in conveying the main narrative, insufficient intensity in generating momentum, limited depth of meaning, and inadequate utilisation of colour. The attractiveness of visual aesthetics is subjective and can vary among individuals. This study introduces a novel approach to address the issues of colorization quality and temporal stability in cinema colorization. The suggested method utilises a generative adversarial network with a recurrent structure, enabling the automatic colorization of films without the need for reference frames or artificial intervention. The network constructs a confrontation network by employing classical conditions. The generator is utilised to produce colour images and fulfil the colorization requirement, while the discriminator is employed to discern the genuineness of the generated images and enhance the performance of the generator. The incorporation of cycle structure and timing consistency loss is proposed as a means to incorporate timing information and address the stability issue in the context of colouring. The experimental findings demonstrate that this particular method is capable of significantly mitigating flickering in the produced movie sequence, while simultaneously preserving the colour integrity of individual frame images. This research article aims to examine and evaluate the visual representation of light and colour in films utilising wavelet neural network techniques.