Research on Compression of Pulsar Candidate Diagnostic Images Based on WBS and Deep Neural Network

The magnitude of candidate diagnostic images generated by FAST pulsar search has increased exponentially, which brings challenges to scientific data management. It is urgent to study compression methods to achieve effective storage of diagnostic images and speed up their transmission and sharing in the network. It is not reasonable to simply treat the whole image as a color image and compress it with the same method. We propose a skip white block coding and a deep network compression coding model to compress pulsar candidate diagnostic images. We take the pulsar candidate diagnostic images from recent FAST sky survey search projects to train and verify the model. The results show that the performance of the improved WBS compression for sparse black and white images is 5 times than PNG. The performance of deep network compression algorithm for PSNR is better than JPEG and JPEG2000, comparable to BPG. And it is much better than traditional compression algorithm for SSIM.