We extend the DISTS index to a locally adaptive version – A-DISTS, which separates structure and texture across space and scale, and assign the adaptive weighting of quality measurements according to local image content.
We performed a large-scale comparison of full-reference IQA models in terms of their use as objectives for the optimization of image processing algorithms — denoising, deblurring, super-resolution, and compression.
We proposed a Deep Image Structure and Texture Similarity (DISTS) metric, which is sensitive to structural distortions, tolerant of texture resampling, and robust to mild geometric transformations.
We presented a social media popularity prediction strategy based on multiple features fusion with deep neural network for ACM Multimedia Challenge.
We conducted a systematic study of intrinsic image popularity assessment on social media, and constructed a large-scale image database, and proposed a DNN-based computational model for popularity prediction.
We presented a robust active contour model driven by local pre-fitting energy for fast image segmentation.
We proposed an active contour model which combines region-scalable fitting energy and optimized Laplacian of Gaussian energy for image segmentation.
We introduced a novel local intensity fitting energy in active contour models to segment the images with intensity inhomogeneity.