Samsung patents method to fix image lighting using neural networks

1. A method, performed by at least one processor, of correcting lighting of an input image, the method comprising: generating, using a first neural network, predicted lighting data corresponding to the lighting of the input image and embedding data corresponding to a feature of the input image, wherein the predicted lighting data is represented as a three-dimensional (3D) vector, with three components representing red color value, green color value, and blue color value of the lighting applied to content of the input image; generating, using a second neural network, a lighting weight corresponding to the input image based on the predicted lighting data, the embedding data, and sensor data, wherein the sensor data indicates an amount of light measured by a sensor, wherein the lighting weight is represented as a single vector including a plurality of lighting weight values; and generating correction lighting data for the input image by applying the lighting weight to preset basis lighting data corresponding to the input image, wherein the preset basis lighting data is determined based on a color temperature of light emitted from a light source or a type of a capturing device that captured the input image, wherein the preset basis lighting data includes a plurality of vectors, and a number of vectors included in the basis lighting data matches the number of lighting weight values of the lighting weight vector, and the generating correction lighting data comprises: determining a first lighting correction vector by calculating a weighted sum between the plurality of vectors of the preset basis lighting data and the plurality of lighting weight values of the lighting weight; and generating the correction lighting data for the input image based on the first lighting correction vector.