Research on Low Resolution Cell Image Feature Fusion Algorithm Based on Convolutional Neural Network

Abstract

In this paper, a low-resolution image fusion method based on convolutional neural network is proposed for the problem of low resolution, few detail features and difficulty in feature extraction of cell images collected by lensless cell acquisition system. Firstly, the image of the cell collected by the medical microscope is segmented into a single white blood cell image with a resolution of 90 × 90 by image threshold segmentation algorithm, and then downsample it to 9 × 9 and input it into the feature fusion network for training. After the training is converged, a feature fusion model is obtained, and then the white blood cell image collected by the lensless cell collection system is input into the model to synthesize the fused cell image with a resolution of 36 × 36. Further, using image binarization and other algorithms, the nucleoplasmic ratio of the fused cell image can be obtained. Finally, the simulated vacuolar white blood cell image with a resolution of 9 × 9 is mixed with the normal white blood cell test image in different proportions and then tested. The test results show that the fused cell image shows a similar topographical feature to the larger part of the mixed test images. This is of great significance for the diagnosis of medical diseases.