Pixel Readout Reordering for Image Sensor Pattern Noise Reduction

The patent segments the readout process by dividing pixels into different groups (e.g., even and odd columns) that are read out at different times. Instead of reading all pixels in a row simultaneously, the readout is split into multiple time slots, which breaks the temporal correlation of noise across the entire row while maintaining high readout throughput through parallel processing of different pixel groups.

The patent implements dynamic ADC assignment where the mapping between columns and ADCs changes over time. Different columns are assigned to different ADCs in different time slots, which distributes the gain variation error across multiple columns rather than concentrating it in single columns. This temporal multiplexing of ADCs reduces the visibility of gain variation as spatial noise while maintaining efficient parallel readout.

The patent applies partial reordering by only reordering specific groups of pixels (e.g., even and odd columns) rather than completely randomizing or fully reordering the entire pixel array. This selective reordering approach achieves sufficient noise reduction by breaking the dominant row-based correlation patterns while keeping the control logic relatively simple and manageable.