Protecting digital images sent over public networks is a genuine and growing concern across healthcare, defence, finance, and personal communication. This paper introduces the Gaussian Key Division (GKD) algorithm — a symmetric encryption scheme where every pixel of a normalised input image is divided element-wise by a randomly generated Gaussian noise matrix, yielding a visually scrambled ciphertext. The mathematical inverse — multiplying by the same matrix — recovers the original with near-perfect accuracy. The full system is built in Python 3.x using OpenCV, NumPy, Pillow, and Tkinter. Tests on five different images show that adjacent-pixel correlation falls from a mean of 0.889 down to virtually zero (0.00002) after encryption, the encrypted PSNR averages 9.54 dB against the original, and decryption reaches a mean PSNR of

62.09 dB with MSE under 0.05 grey levels. The paper openly addresses the algorithm's known weaknesses — low ciphertext entropy and the lack of a diffusion layer — positioning GKD as a practical, lightweight obfuscation tool suited to learning environments and non-adversarial deployments.