Coordinated-axis Weight for Real-Time Trace Transform

This paper presents a method for building Coordinated-axis Weight (CW) that has the advantage can be used to increase the speed for converting image data into the domain of real-time trace transforms. The main idea for this method is to use a feature reduction method for image data that must be used in cooperation with supervised machine learning outcome predictions. But the obtained result still retains the distinctive features of the data image in the form of a trace transform. Start by reducing the number of features from the image data, which will be used to reduce the size of the image to the appropriate image size, which in this article will reduce the image size to 45x45 pixels. The next step is to train the data using a Decision Tree (DT) method by aligning the reduced image data in a 1D model. In addition to the image data, two more features are added, i.e. the
$\phi$
and
$\rho$
positions, to inform the prediction that each position has a different value. And the last step is to bring the data to create a model for predicting the answer. The weights obtained from the model are called Coordinated-axis Weights (CW). The experiments within this paper were experimented using the Caltech-256 database, which was able to measure the MSE, RMSE, and PNSR of the images compared to the original trace transform response images at 4.1, 2.02, and 42.00, respectively, including the highest predictive speed of 0.2 ms per image.