Abstract
A STUDY ON THE METHODS OF IMAGE FEATURE EXTRACTION
*Sayeed Mohammed Shaik and Dr. R. Kamalraj
ABSTRACT
Computer Vision tasks analyze the whole image. These tasks are computationally expensive. By extracting features from the image, we can reduce the computational expense. Using Deep Learning we can create Intelligent Models that can perform these Computer Vision tasks. We can tackle each task such as image classification and object detection with Models specifically trained to deal with those tasks. By combining said specifically trained Models we can achieve an aggregated output. Some of the techniques such as Instance Segmentation deal with the Computer Vision tasks in similar fashion. Features in deep learning can range from luminance of the pixels to high level pattern classification. Images are represented as an array of pixels, we can utilize Convolutional Neural Networks for the Image Processing removing the need to transform the array of pixels. Other Neural Network architectures such as U-net and Auto-Encoders can also be utilized in dealing with similar Computer Vision tasks. Many Models are trained to deal with specific tasks. This paper surveys some of the Models currently performing the Computer Vision tasks and Image processing.
[Full Text Article] [Download Certificate]