Multi-Exposure Image Fusion using Convolutional Neural Network (CNN)

Because calculating the amount of light in a photograph has been the topic of so much research, multiexposure photos frequently have a smeared appearance. The fundamental purpose of these investigations is to identify the patch-smart solution for illumination estimation with human intervention using the CNN method, which can be utilized to distinguish between distinct time periods in multi-exposure fusion photos. Common picture fusion segmentation techniques, such as the seeded watershed algorithm, are time-consuming and error-prone. In addition, parameter adjustment is necessary when employing such algorithms, therefore a competent consumer is desirable. This research intends to address the critical issues connected with patch-wise illumination estimation in multi-source photo fusion. Using convolutional neural networks, multiple-publicity image fusion snapshots will be created. The neural network can be trained with visual data that has been manually tagged, and its performance can be measured with summative metrics.