Sports Training Image Blur and Deduplication Algorithms Based on Visual Sensing

The Super-Resolution Convolutional Neural Network (SRCNN) was employed in this study to enhance the clarity and minimize blur in sports training photographs. High-resolution cameras were utilized to capture sports training images and perform denoising on the images. To improve the nonlinear expression capabilities, the SRCNN model and activation function were introduced after the input image was reduced in size using downsampling techniques. Upsampling was used to transform the feature map back into a high-resolution image. The experimental results showed that the average peak signal-to-noise ratio (PSNR) of SRCNN was 93dB, and its average structural similarity (SSIM) index was 0.94. SRCNN can be used to produce clearer, higher-quality sports training photos.