Image Noise and Noise Images in Computer Vision, Photography and Embedded Vision

What is Image Noise?

Image Noise is the variation in signal  from the ground trust, from pixel to pixel or frame to frame.

How Does A Noisy Image Impact Human Viewing and Computer Vision?

Noise is commonly thought of as Salt-and-Pepper noise. In a still image, one may say the image appears to be "Too Grainy"

In this sailboat example, researchers show that the confidence of class recognition drops rapidly  multi-pixel / spatial noise is present in an image.

Dodge et al. "Understanding How Image Quality Affects Deep Neural Networks"  

What Types of Noise Exist in Cameras?

There are numerous types of noise in cameras. There are two main categories:

• Single pixel noise can be thought of as temporal or random. Single Pixel noise is a variation in signal (or pixel value) from frame to frame. 
• Multi-pixel noise can be thought of as spatial or fixed pattern noise, with variation from pixel to pixel in a single frame. 

The specific types of noise in a camera system are:

• Readout Noise
• Photon Shot Noise
• Fixed Pattern Noise
• Reset Noise

What About Denoising?

Denoising can be applied to mitigate noise. The approach is different for each of the two categories of noise.

Many multi-pixel (Spatial) denoising algorithms use a frequency filter to reduce the variation in pixel intensity. This filter also filters out information included in the image.

Hasinoff, et. al. "Burst Photography for high dynamic range and low light cameras"

Denoising (Spatial) Creates Texture Blur and Loss 

Localized denoisingcreates texture loss, which is particularly visible with mobile phone cameras.

This texture loss is noticeable to the human visual system when zooming in on the image.

For computer vision, the impact can be even more noticeable depending on your filter size.

This example by Chen et al shows a brick wall building and the output from an edge detection network before and after filtering. The high frequency noise filter greatly reduces the texture in the image, particularly in regions where color is constant.

Chen, S.-C.; Chiu, C.-C. Texture Construction Edge Detection Algorithm. Appl. Sci. 2019, 9, 897. https://doi.org/10.3390/app9050897

References and Related Links

Dodge et al. "Understanding How Image Quality Affects Deep Neural Networks" 

Stanford EE 392B lecture notes