Scene classification for image processing

A method comprising:

• obtaining, from an image sensor, by an image signal processor, a first input image;
• obtaining, by the image signal processor, color component ratio data corresponding to respective pixels from the first input image;
• obtaining pixel classification data for the respective pixels using the color component ratio data;
• obtaining scene classification data indicating whether a scene captured by the first input image is an underwater scene using the pixel classification data;
• and
• outputting processed image data for the first input image using the scene classification data.

An apparatus comprising:

• an image sensor;
• a processor configured to:
• obtain a first input image from the image sensor;
• obtain color component ratio data corresponding to respective pixels from the first input image;
• obtain pixel classification data for the respective pixels using the color component ratio data;
• obtain scene classification data indicating whether a scene captured by the first input image is a nighttime scene using the pixel classification data, wherein to obtain the scene classification data the processor is configured to:
• obtain a respective cardinality of pixels corresponding to respective candidate scene classifications using the pixel classification data for the respective pixels;
• and
• obtain the scene classification data using the respective cardinality of pixels corresponding to respective candidate scene classifications;
• and
• output processed image data for the first input image using the scene classification data.