Noise
Random variations in brightness or color unrelated to the actual scene content. Becomes prominent in high-ISO photography and long exposures.
Image noise refers to random fluctuations in pixel values that do not correspond to actual scene information. It is the digital equivalent of film grain and arises from the physical properties of image sensors and their electronic readout circuits. Noise obscures fine detail and degrades perceived image quality, particularly in shadow regions and uniform areas.
- Shot noise (photon noise): Statistical variation inherent to the quantum nature of light. More noticeable in darker regions where fewer photons are captured
- Thermal noise (dark current): Electronic noise that increases with sensor temperature. Becomes significant during long exposures or in hot environments. Astrophotographers subtract dark frames to compensate
- Color vs. luminance noise: Color (chrominance) noise manifests as random colored speckles, while luminance noise appears as brightness grain. Some noise reduction tools process these components separately
Denoising techniques range from classical filters (Gaussian, median, bilateral) to deep learning approaches (DnCNN, NAFNet). In programmatic workflows, OpenCV provides cv2.fastNlMeansDenoisingColored() for non-local means denoising.