Image preprocessing with bi-histogram equalization in a supervised classification of radiographs


Convolutional neural network models are known for their effectiveness in solving computer vision problems. These neural networks are designed to extract features from raw data for accurate classification. The objective of this paper is to study the effects of bi-histogram equalization preprocessing on the performance of a convolutional neural network in identifying pneumonia from a chest x-ray image. The model trained with the preprocessed images demonstrated improved accuracy during training and validation of the neural network.