Quaternion Convolutional Neural Networks for Heterogeneous Image Processing

Abstract

Convolutional neural networks (CNN) have recently achieved state-of-the-art results in various applications. In the case of image recognition, an ideal model has to learn independently of the training data, both local dependencies between the three components (R,G,B) of a pixel, and the global relations describing edges or shapes, making it efficient with small or heterogeneous datasets. Quaternion-valued convolutional neural networks (QCNN) solved this problematic by introducing multidimensional algebra to CNN. This paper proposes to explore the fundamental reason of the success of QCNN over CNN, by investigating the impact of the Hamilton product on a color image reconstruction task performed from a gray-scale only training. By learning independently both internal and external relations and with less parameters than real valued convolutional encoder-decoder (CAE), quaternion convolutional encoder-decoders (QCAE) perfectly reconstructed unseen color images while CAE produced worst and gray-scale versions.

arXiv commentSubmitted at ICASSP 2019

References in corpus (4)

• Adam: A Method for Stochastic Optimization
• A guide to convolution arithmetic for deep learning
• Lossy Image Compression with Compressive Autoencoders
• Quaternion Recurrent Neural Networks

Cited by in corpus (21)

• Quaternion-valued Correlation Learning for Few-Shot Semantic Segmentation
• PHNNs: Lightweight Neural Networks via Parameterized Hypercomplex Convolutions
• Beyond Fully-Connected Layers with Quaternions: Parameterization of Hypercomplex Multiplications with $1/n$ Parameters
• Quaternion Generative Adversarial Networks
• A general framework for constrained convex quaternion optimization
• Color Image Inpainting via Robust Pure Quaternion Matrix Completion: Error Bound and Weighted Loss
• Hypercomplex-Valued Recurrent Correlation Neural Networks
• Compressing deep quaternion neural networks with targeted regularization
• Quaternion Convolutional Neural Networks: Current Advances and Future Directions
• Parameterized Hypercomplex Graph Neural Networks for Graph Classification
• Quaternion-Valued Convolutional Neural Network Applied for Acute Lymphoblastic Leukemia Diagnosis
• Hypercomplex Image-to-Image Translation
• Generalizing Medical Image Representations via Quaternion Wavelet Networks
• QuatDE: Dynamic Quaternion Embedding for Knowledge Graph Completion
• Traffic flow prediction using Deep Sedenion Networks
• Generalizing Complex/Hyper-complex Convolutions to Vector Map Convolutions
• A Quaternion-Valued Variational Autoencoder
• Quaternion Neural Networks for Multi-channel Distant Speech Recognition
• Quaternion Factorization Machines: A Lightweight Solution to Intricate Feature Interaction Modelling
• Speech Emotion Recognition System by Quaternion Nonlinear Echo State Network
• RingCNN: Exploiting Algebraically-Sparse Ring Tensors for Energy-Efficient CNN-Based Computational Imaging