Canny edge detection for blurred textures in advertising art images based on generative adversarial networks

Abstract For high-dynamic-range display devices, multispectral imaging environments, and non-uniform illumination acquisition systems, advertising art images are prone to complex degradations such as optical chromatic aberration diffraction, sensor readout noise, and motion blur coupling. These issues cause traditional edge detection methods to suffer from high-frequency texture annihilation and semantic edge discontinuity. This paper proposes a Canny edge detection method for blurred textures in advertising art images based on generative adversarial networks (GANs). Advertising art images are captured, and their blur-invariant features are extracted using local phase quantization techniques. Combined with a generative adversarial network, this approach achieves image restoration and edge enhancement. A hybrid filtering technique—combining adaptive median filtering and guided filtering—replaces the traditional Gaussian filter in Canny edge detection to improve noise reduction and edge preservation capabilities. Additionally, multi-directional Sobel operators and an adaptive threshold mechanism are introduced to achieve precise edge extraction. Experiments demonstrate that under low-clarity advertising art images, this method achieves edge localization errors below 1.72 pixels, edge integrity exceeding 90%, and recognition error rates below 5.6%. Peak signal-to-noise ratio values remain consistently above 30 dB, while computational efficiency is maintained between 1.3 and 1.5 s. This significantly enhances the accuracy and real-time performance of edge recognition in advertising art images, providing a reliable technical foundation for intelligent image analysis and semantic understanding.