An efficient coding‐based grayscale image automatic colorization method combined with attention mechanism

Abstract

The development of deep learning provides a new way for solving the colorization problem on the grayscale image. Excellent coding-based methods appear in the automatic image colorization task, avoiding the unsaturated colour effect problem of previous methods based on the L2 loss function. Traditional neural networks come with high computational costs and a large number of parameters. Considering the limitation of memory and computing resources and aiming at lightweight, a novel grey image automatic colorization network is proposed. The basic idea of coding-based methods is used, regarding the colorization task as a pixel-level classification problem, meanwhile redesign and improve the colour encoding and decoding process. This network architecture leverages a lightweight convolution to reduce the computation and combines an efficient attention model to form a residual block as the kernel of the backbone network. Furthermore, an efficient image self-attention mechanism placed at the end of the network is applied to enhance the ultimate colouring results. The method proposed in this paper can maintain the natural colouring effect and significantly reduce the computational amount and network model parameters.

Interactive image colorization method via dual local self-attention

“…This method demonstrates the effectiveness of CNN architectures in solving image colorization without the need for reference images or manual interaction. Qin et al [6] discussed an image colorization method using coding. This method is more efficient, saves memory, and achieves natural shading effects.…”

“…This integration of model-based approaches with human expertise and artistic intuition enhances the quality and freedom of colorization, thereby assisting professionals in executing more precise and creative colorization. In existing research methods [4][5][6][7][8][9], to expand the model's receptive field and propagate color cues to distant regions awaiting colorization, a common approach involves the design of heavily stacked Convolutional Neural Networks (CNNs). The color information of large semantic areas is only propagated in deep CNNs, and the transitional connections of color in spatial areas are often ignored.…”