Image Quality Enhancement and Caption Generation
Kumarasamy Saravanan, S SwethaSri, M Swetha, N Renuka · 2023
Our brain is capable of labeling or annotating any image that is visible to us. But the machine does not have the capability of thinking on its own. To bringing it possible, we are employing Deep Learning Techniques. The aims of image captioning are to produce natural language expressions, to capture the relationship between the objects in the image, and to evaluate the credibility of the constructed descriptions. Uprooted features of the image are handled by Xception. Xception is a CNN model. These extracted features will be permitted by the LSTM model, and it will allow for the image subtitle. When the descriptions provided are just one word long, the object detection task for image captioning becomes more generic. But the problem arises when the image is of poor quality. The Images taken by a user under various lighting circumstances, Dots Per Inch (DPI), and different angles may have worse quality. We are utilizing several techniques, object identification, image classification, CNN, and OCR, to enhance the quality of images.Reducing Noise, Rescaling, Binarization / Thresholding, Removing Skewness / Deskew, and Morphological Operations are a few actionstaken during the image quality enhancement process.