Real-Time Film Effect Image Processing for Video Capture

The patent segments the image processing task into multiple processing stages (e.g., edge detection, skin tone adjustment, detail enhancement) that can be executed sequentially or in parallel. This allows the system to apply complex deep learning operations only to specific regions or aspects of the image rather than processing the entire image with full deep learning complexity, thereby reducing overall processing time while maintaining image quality.

The patent employs preliminary actions by pre-processing images to extract key features (such as skin regions, edges, and important structures) before applying deep learning-based enhancement. This pre-extraction of relevant information reduces the computational burden during the actual enhancement phase, enabling faster processing while preserving the ability to achieve high-quality results.

The patent applies local quality by differentiating processing strategies based on image characteristics. Complex deep learning-based operations are applied selectively to specific regions (such as skin areas or regions with important details) rather than uniformly to the entire image. This localized approach maintains high quality where needed while significantly reducing overall power consumption.

The patent implements partial action by applying full deep learning processing only to a subset of images or specific regions within images, rather than applying exhaustive processing to all images. This selective application of processing power reduces energy consumption while maintaining acceptable quality levels across the overall output.