As technologies for image processing, image enhancement can provide more effective information for later data mining and image compression can reduce storage space. In this paper, a smart enhancement scheme during decompression, which combined a novel two-dimensional F-shift (TDFS) transformation and a non-standard two-dimensional wavelet transform (NSTW), is proposed. During the decompression, the first coefficient s00 of the wavelet synopsis was used to adaptively adjust the global gray level of the reconstructed image. Next, the contrast-limited adaptive histogram equalization (CLAHE) was used to achieve the enhancement effect. To avoid a blocking effect, CLAHE was used when the synopsis was decompressed to the second-to-last level. At this time, we only enhanced the low-frequency component and did not change the high-frequency component. Lastly, we used CLAHE again after the image reconstruction. Through experiments, the effectiveness of our scheme was verified. Compared with the existing methods, the compression properties were preserved and the image details and contrast could also be enhanced. The experimental results showed that the image contrast, information entropy, and average gradient were greatly improved compared with the existing methods.
18 May 2021 in Proceedings of the AAAI Conference on Artificial Intelligence
Bilinear pooling has achieved an excellent performance in many computer vision tasks such as fine-grained classification, scene recognition and texture recognition. However, the high-dimension features from bilinear pooling can sometimes be inefficient and prone to over-fitting. Random Maclaurin (RM) is a widely used GPU-friendly approximation method to reduce the dimensionality of bilinear features. However, to achieve good performance, large projection matrices are usually required in practice, making it costly in computation and memory. In this paper, we propose a Shifted Random Maclaurin (SRM) strategy for fast and compact bilinear pooling. With merely negligible extra computational cost, the proposed SRM provides an estimator with a provably smaller variance than RM, which benefits accurate kernel approximation and thus the learning performance. Using a small projection matrix, the proposed SRM achieves a comparable estimation performance as RM based on a large projection matrix, and thus boosts the efficiency. Furthermore, we upgrade the proposed SRM to SRM+ to further improve the efficiency and make the compact bilinear pooling compatible with fast matrix normalization. Fast and Compact Bilinear Network (FCBN) built upon the proposed SRM+ is devised, achieving an end-to-end training. Systematic experiments conducted on four public datasets demonstrate the effectiveness and efficiency of the proposed FCBN.
24 July 2019 in Health Information Science and Systems