GaussianImage is a novel paradigm for image representation and compression that leverages 2D Gaussian Splatting. This approach offers a discrete and explicit representation of images, significantly reducing computational complexity and GPU memory usage compared to implicit neural representations (INRs). The method uses 2D Gaussians, each defined by position, anisotropic covariance, color coefficients, and opacity, resulting in a 6.5× compression ratio over 3D Gaussians. A novel accumulated blending-based rasterization algorithm replaces depth-based sorting and alpha-blending, improving training efficiency and rendering speed. The method achieves 1500-2000 FPS rendering speed and competitive rate-distortion performance with minimal GPU memory usage. The authors also develop a low-complexity neural image codec using vector quantization, achieving rates comparable to COIN and COIN++ with partial bits-back coding. Experimental results demonstrate the effectiveness of GaussianImage in image representation and compression, making it a promising technique for efficient and high-quality image processing.GaussianImage is a novel paradigm for image representation and compression that leverages 2D Gaussian Splatting. This approach offers a discrete and explicit representation of images, significantly reducing computational complexity and GPU memory usage compared to implicit neural representations (INRs). The method uses 2D Gaussians, each defined by position, anisotropic covariance, color coefficients, and opacity, resulting in a 6.5× compression ratio over 3D Gaussians. A novel accumulated blending-based rasterization algorithm replaces depth-based sorting and alpha-blending, improving training efficiency and rendering speed. The method achieves 1500-2000 FPS rendering speed and competitive rate-distortion performance with minimal GPU memory usage. The authors also develop a low-complexity neural image codec using vector quantization, achieving rates comparable to COIN and COIN++ with partial bits-back coding. Experimental results demonstrate the effectiveness of GaussianImage in image representation and compression, making it a promising technique for efficient and high-quality image processing.