The scan angle of a resonant scanner over time follows a sinusoidal pattern, so that the angular velocity of the scanner is greatest at 0° deflection, while the angular velocity is 0 at the maximum deflection. If the non-uniform angular velocity of the scanner is ignored during image formation, the resulting image is warped in x. To form pixels, ScanImage averages multiple digitizer samples using a FPGA. Since the digitizer samples at a constant rate, a varying number of samples needs to be used for each pixel to correct for the non-uniform angular velocity of the scanner. ScanImage calculates a sample to pixel mask at the start of an acquisition and stores it on the FPGA. The FPGA then applies the mask to the incoming data stream from the digitizer to form a line.
Since more samples are averaged at the edges of the field of view and the edge of the field of view is scanned slower than the center, the signal to noise ratio varies over the field of view. This can easily be observed by imaging a uniform sample such as a fluorescent slide: The center of the field of view (in x) will show a higher SNR then the edges of the field of view.
Mask used by ScanImage for a typical acquisition (pixels per line: 512, spatial fillfraction: 0.9, resonant frequency: 7910Hz, 80MHz sample rate)
Example of samples to pixel assignment for a line with 8 pixels, and a spatial fillfraction of 0.8. The part of the scan used for image formation is highlighted in green.
To achieve a uniform scan angle per pixel, a varying number of digitizer samples are used to form the pixels. Note that the assignment is imperfect because of quantization errors.
In ScanImage, this sample to pixel mask is calculated for every acquisition, and applied to the input data stream on the FPGA. A pixel is formed by calculating the arithmetic mean of the number of samples specified in the sample to pixel mask. Note: because the algorithm is implemented on an FPGA, the mean uses integer division.
