Forgery detection method based on feature enhancement and spectral analysis

The invention provides a counterfeiting detection method based on feature enhancement and spectral analysis, and the method comprises the steps: retaining texture information through shallow texture feature enhancement in an RGB space, and obtaining a texture feature enhancement graph; an attention map is obtained in an RGB space through a channel space attention mechanism module, element-by-element point multiplication is carried out on the attention map and the input feature map, and a deep semantic feature enhancement map after attention enhancement is obtained; the method comprises the following steps: converting an image from an RGB color space to a YCbCr color space through color space conversion in a frequency domain space, connecting different frequency domain channels in series through DFT and DWT frequency domain conversion to obtain a shallow frequency domain feature, and obtaining a deep frequency domain feature map through a feature extraction network; fusing the textural feature enhancement grap