Texture-based Error Analysis for Image Super-Resolution.

Evaluation practices for image super-resolution (SR) use a single-value metric, the PSNR or SSIM, to determine model performance. This provides little insight into the source of errors and model behavior. Therefore, it is beneficial to move beyond the conventional approach and reconceptualize evaluation with interpretability as our main priority. We focus on a thorough error analysis from a variety of perspectives. Our key contribution is to leverage a texture classifier, which enables us to assign patches with semantic labels, to identify the source of SR errors both globally and locally. We then use this to determine (a) the semantic alignment of SR datasets, (b) how SR models perform on each label, (c) to what extent high-resolution (HR) and SR patches semantically correspond, and more. Through these different angles, we are able to highlight potential pitfalls and blindspots. Our overall investigation highlights numerous unexpected insights. We hope this work serves as an initial step for debugging blackbox SR networks.

Dynamic High-Pass Filtering and Multi-Spectral Attention for Image Super-Resolution.

Deep convolutional neural networks (CNNs) have pushed forward the frontier of super-resolution (SR) research. However, current CNN models exhibit a major flaw: they are biased towards learning low-frequency signals. This bias becomes more problematic for the image SR task which targets reconstructing all fine details and image textures. To tackle this challenge, we propose to improve the learning of high-frequency features both locally and globally and introduce two novel architectural units to existing SR models. Specifically, we propose a dynamic highpass filtering (HPF) module that locally applies adaptive filter weights for each spatial location and channel group to preserve high-frequency signals. We also propose a matrix multi-spectral channel attention (MMCA) module that predicts the attention map of features decomposed in the frequency domain. This module operates in a global context to adaptively recalibrate feature responses at different frequencies. Extensive qualitative and quantitative results demonstrate that our proposed modules achieve better accuracy and visual improvements against state-of-the-art methods on several benchmark datasets.

Channel Embedding for Informative Protein Identification from Highly Multiplexed Images.

Interest is growing rapidly in using deep learning to classify biomedical images, and interpreting these deep-learned models is necessary for life-critical decisions and scientific discovery. Effective interpretation techniques accelerate biomarker discovery and provide new insights into the etiology, diagnosis, and treatment of disease. Most interpretation techniques aim to discover spatially-salient regions within images, but few techniques consider imagery with multiple channels of information. For instance, highly multiplexed tumor and tissue images have 30-100 channels and require interpretation methods that work across many channels to provide deep molecular insights. We propose a novel channel embedding method that extracts features from each channel. We then use these features to train a classifier for prediction. Using this channel embedding, we apply an interpretation method to rank the most discriminative channels. To validate our approach, we conduct an ablation study on a synthetic dataset. Moreover, we demonstrate that our method aligns with biological findings on highly multiplexed images of breast cancer cells while outperforming baseline pipelines. Code is available at https://sabdelmagid.github.io/miccai2020-project/.