Low exposure image frame generation algorithms for feature extraction and classification

摘要

Neuromorphic architectures enable machine learning on a faster timescale compared to conventional processorsand require encoding of spike trains from images for computer vision applications. We report low-exposureimage representation algorithms that can generate multiple short-exposure frames from a given long exposureimage. The frame deconvolution is non-linear in the sense that the difference between adjacent short-exposureframes change with exposure time, however the frames have a structural representation of the original imagesuch that the image reconstructed from these frames has a Peak Signal-to-Noise Ratio (PSNR) of over 300 anda Structural Similarity Index Metric (SSIM) close to unity. We show that the low-exposure frames generatedby our algorithms enable feature extraction for machine learning or deep learning, e. g., classification usingconvolutional neural networks. The validation accuracy for classification depends on the range of the randomsubtraction parameter, a used in our algorithms to simulate low-exposure frames. When the maximum of a,equals to the largest allowed change in the pixel intensity per time step, the validation accuracy for classificationof digits in the Digits dataset is 90±3% based on the 1st 1 ms frame. The accuracy increases to 97% with only40% of the 1ms frames generated for a given exposure time. These results show that machine learning can beextended to low exposure images.