This paper explores the application of “color” elements by combining SD and GoogLeNet models to build a dataset of color emotion imagery for film and television, generating a color emotion adversarial network based on conditional convolution, and designing a deep learning-based color emotionalization method. A feature extraction method is used to explore the application of graphic elements in digital effects films and television. Color features were extracted using the color histogram and color moments, the texture was extracted using the CLBP algorithm, and EEG signals were extracted using the CSP method. After analysis, it is concluded that as the film and TV timeline progresses, the value of RGB of color increases from (10,200) to (30,250), and the hue becomes heavier to pave the way for the restless and negative emotions later. The primary effects of graphical emotions were significantly smaller than 0.05 and were capable of causing emotional changes.