Automatic image quality parameter tuning for user-selected visual characteristics based on machine learning

An apparatus comprising:

• an interface configured to receive pixel data for one or more raw images;
• and
• a processor configured to process said pixel data arranged as video frames and perform computer vision operations on said video frames, said processor comprising (i) a digital signal processing pipeline configured to generate processed image data based on said pixel data of said one or more raw images and a set of image quality parameters, and (ii) a tuning tool configured to generate said set of image quality parameters based on said pixel data of said one or more raw images by executing a first artificial neural network model, wherein
• (a) said first artificial neural network model was trained using a set of said raw images, a set of reference images, and a loss function, and
• (b) said image quality parameters relate to one or more of black level correction, lens distortion correction, sensor defect detection, spatial and temporal noise reduction, Bayer demosaicing, color and tone correction, brightness and saturation, contrast, sharpness, white balance, autofocus statistics, auto-exposure statistics, automatic white balance statistics, lens shading correction, color space conversion, gamma correction, dynamic range, electronic image stabilization, decompanding, anti-aliasing, chromatic aberration, digital gain, vignette compensation, and statistics extraction.

A method of tuning image quality parameters of a digital signal processing pipeline comprising:

• receiving one or more raw images and one or more reference images;
• generating a set of image quality parameters based on said one or more raw images by executing a first artificial neural network model;
• generating processed image data by applying said set of image quality parameters on said one or more raw images using said digital signal processing pipeline;
• generating a loss signal based on said processed image data, said one or more reference images, and a loss function by executing a second artificial neural network model;
• adjusting one or more internal neural network parameters of said first artificial neural network model based on said one or more raw images, said one or more reference images, and said loss signal;
• and
• tuning is repeated until said loss signal reaches a predetermined threshold.