Imaging Device Fixed Pattern Noise Estimation Using Partial Pixel Readout

The patent divides the pixel array into multiple regions and reads noise data only from specific regions (e.g., first and second regions) rather than the entire screen. This segmentation approach reduces the amount of noise data that needs to be read and processed, thereby improving continuous shooting speed while still maintaining effective fixed pattern noise elimination through the use of read noise data in combination with stored fixed pattern noise.

The patent performs preliminary action by storing fixed pattern noise data for multiple screens in advance in a buffer memory during periods when continuous shooting is not occurring or during idle time. This allows the system to have pre-prepared noise data ready for use, eliminating the need to read and process noise data in real-time during continuous shooting operations, thus maintaining high shooting speeds while still achieving accurate noise elimination.

The patent extracts only the essential noise data from specific regions of the pixel array rather than reading and storing noise data from the entire screen. By taking out only the necessary portions of noise data and combining them with previously stored fixed pattern noise, the system reduces buffer memory usage and avoids memory saturation, thereby maintaining high continuous shooting speeds while still achieving effective noise elimination.

The patent uses a copying approach by utilizing stored fixed pattern noise data from previously captured screens and combining it with newly read noise data from specific regions. This creates a virtual representation of comprehensive noise data without physically storing all the raw noise data for every screen in the buffer memory. The system effectively copies and reuses noise patterns across multiple frames, reducing memory requirements while maintaining noise elimination capability.