Image rotation estimation by angle scanning and channels average method in optical correlator

Due to the large storage capacity and the high speed multi-channel parallel correlation, the holographic optical correlator (HOC) is especially suitable for the on-board real time image recognition. However, when the target image contains rotation to the template images, the rotation needs to be compensated before the translation measurement. It is still a challenging task for the traditional optical correlator. A novel approach, which combines the multi-angle inputting of a rotated target image and the multi-channel averaging of the correlation outputs, is proposed to measure the rotation. The target image is ON–OFF pixel inverted to adapt to the linear working range of the CCD. Then the inverted image is rotated at a small angle intervals sequentially. The several generated rotated images are uploaded into the HOC to correlate with all the template images in parallel. The correlation value for each input image is based on the average of intensities of the neighbor template images. It is proved that the minimum of all the correlation values could determine the rotation angle value of the original target image. The experiments are performed with the remote sensing images and the measurement accuracy is within ±0.14° at the rotation range of ±6°.