Improving Machine Learning Based Color Optimization Efficiency by Using a New Image Restoration Technique

Abstract

Listen

Translate article

In the past ten years, neural network is playing an important role of solving computer vision problems. One reason is due to the improvement of machine learning algorithms. The other reason is due to the hardware acceleration from GPUs and TPUs. However, researchers and developers are still having a hard time to deal with high resolution images. While the image size increases, the computational time of neural network models may increase quadratically. In this paper, we introduced a new image restoration method which successfully improves the efficiency of image color optimization task. For any input image size, this method could reduce the computational time of the autoencoder to a constant time.