Spatial Moment Pooling Improves Neural Image Assessment

TL;DR

This paper introduces spatial moment pooling (SMP), extending the common spatial average pooling in CNNs for blind image quality assessment, leading to improved performance by capturing higher order spatial statistics.

Contribution

The paper proposes a novel spatial moment pooling method that incorporates higher order moments into CNN features for blind IQA, enhancing accuracy over traditional average pooling.

Findings

SMP significantly improves CNN-based blind IQA performance.

The method achieves state-of-the-art results in image quality assessment.

Normalization techniques enable stable training with higher moments.

Abstract

In recent years, there has been widespread attention drawn to convolutional neural network (CNN) based blind image quality assessment (IQA). A large number of works start by extracting deep features from CNN. Then, those features are processed through spatial average pooling (SAP) and fully connected layers to predict quality. Inspired by full reference IQA and texture features, in this paper, we extend SAP ($1^{st}$ moment) into spatial moment pooling (SMP) by incorporating higher order moments (such as variance, skewness). Moreover, we provide learning friendly normalization to circumvent numerical issue when computing gradients of higher moments. Experimental results suggest that simply upgrading SAP to SMP significantly enhances CNN-based blind IQA methods and achieves state of the art performance.