Mean squared error: Love it or leave it? A new look at Signal Fidelity Measures

2009

779 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition, Evaluation & Benchmarks, Signal Processing

Citations in the corpus

• A note on the evaluation of generative models (2015, neutral): “But already perceptually small changes can lead to large changes in Euclidean distance, as is well known in the psychophysics literature (e.g., Wang & Bovik, 2009).”
• Perceptual Losses for Real-Time Style Transfer and Super-Resolution (2016, neutral): “The traditional metrics used to evaluate super-resolution are PSNR and SSIM [54], both of which have been found to correlate poorly with human assessment of visual quality [55,56,57,58,59].”
• Autoencoding beyond pixels using a learned similarity metric (2015, neutral): “A more detailed discussion on the subject is provided by Wang & Bovik (2009).”
• Survey on Automated Machine Learning (2019, neutral): “In general, the loss function depends on the underlying problem: a common loss function for regression is the mean squared error (Wang and Bovik, 2009) and hinge loss is often used for classification (Rosasco et al., 2004).”
• Adversarial Examples Are a Natural Consequence of Test Error in Noise (2019, positive): “In order to visualize worst-case perturbations at varying l 2 distances, we visualize an image that minimizes similarity according to the SSIM metric (Wang & Bovik, 2009).”
• Pros and Cons of GAN Evaluation Measures (2018, neutral): “This is a well known phenomenon in the psychophysics literature (See Wang and Bovik [88]).”
• fastMRI: An Open Dataset and Benchmarks for Accelerated MRI (2018, neutral): “The assessment of MRI reconstruction quality is of paramount relevance to develop and compare machine learning and medical imaging systems [51, 53, 3, 58].”
• Group MAD Competition? A New Methodology to Compare Objective Image Quality Models (2016, neutral): “In recent years, there has been a rapidly growing interest in developing objective image quality assessment (IQA) models that can automate the process [30, 31].”
• Loss Functions for Image Restoration With Neural Networks (2017, positive): “Moreover, given that it can be used as a differentiable cost function, SSIM has also been used in iterative algorithms designed for image compression [16],”
• Loss Functions for Neural Networks for Image Processing (2015, neutral): “There is longer list of reasons for which we refer the reader to the work of Wang and Bovik [16].”
• No-Reference Image Quality Assessment in the Spatial Domain (2012, neutral): “Although PSNR is a poor measure of perceptual quality, it is often used to benchmark for QA algorithms [41], [42].”
• Disruptive Innovations and Disruptive Assurance: Assuring Machine Learning and Autonomy (2019, neutral): “13 Finally, adversarial perturbations can be achieved by much simpler attacks that do not require ML algorithms (e.g., covering a stop sign).”
• From QoS to QoE: A Tutorial on Video Quality Assessment (2015, neutral): “Structural Similarity Based Models: The objective of the structural similarity based models is to measure the similarity (fidelity) between the original video and the distorted video, based on the knowledge of the transmitter, channel and the receiver [109].”
• Deep Learning for Image Super-Resolution: A Survey (2019, positive): “Since the SSIM evaluates the reconstruction quality from the perspective of the HVS, it better meets the requirements of perceptual assessment [60], [61], and is also widely used.”
• Hyperspectral Image Denoising Employing a Spectral–Spatial Adaptive Total Variation Model (2012, neutral): “Recently, some image quality assessment indices based on the human vision system have been proposed, such as the structural similarity (SSIM) index [43], [44] and the multiscale geometric analysis-based indices [45], [46].”