Image Processing Methods: A Systematic Literature Review of Classical and Modern Approaches

Abstract

In image processing, there are many different methods used to process pictures. This includes denoising, enhancement, segmentation, feature extraction, and classification. All of them join together to solve different problems and understand the changes in images. They are useful in many ways, like in medicine, security, photography, and robotics. Where images need to be studied or improved. Drawing on visual information, these methods help us in comprehending images, extracting key data, and making informed choices. There are two main ways to process the image, which are through traditional image-processing methods and deep-learning models. Usually, traditional techniques depend on manually designed algorithms and rules. Which uses fixed steps to process images. In contrast, deep learning models learn features directly from the information itself, enabling them to automatically detect distant details that traditional techniques could miss.  The things that help the image processing methods to proceed are like Self2Self NN, Denoising DFT-Net CNNs, and MPR-CNN, which help remove unwanted noise from images in denoising. However, they still face difficulties with data preparation and adjusting model settings. While in an image enhancement, R2R and LE-net are employed to enhance the image’s visual quality, through which they can deal with complex real-world images and help them to look natural. On one hand, in the segmentation, PSP Net and Mask-RCNN methods accurately separate objects in an image; however, they can face problems with overlapping objects and ensuring reliable performance. In the method of feature extraction, models like CNN and HLF-DIP can automatically detect important image details, though they can be hard to interpret and sometimes complex to use. In the classification method, Residual Networks and CNN-LSTM are the approaches, which are effective at accurately identifying image categories; however, they require enormous computing power and can be difficult to fully understand. This review gives a clear overview of the advantages and disadvantages of different methods, which can help people choose the best approach for real-world use. As image processing continues to develop, solving problems like high computing needs and ensuring reliable performance will be important to make these techniques work at their best.