Image Texture Generation Using Autoregressive Integrated Moving Average (ARIMA) Models.
Abstract
This thesis involves investigation of linear filtering models as a means of generating texture in images. Various autoregressive filter models are used to generate various textures, and the results are analyzed to determine relationships between filter parameters and texture characteristics. A two-dimensional counterpart to the autoregressive integrated moving average (ARIMA) model from one-dimensional time series analysis theory is developed and tested for texture modeling applications. All these models are driven by white noise, and to the extent that real images can be reproduced this way, advantages in image texture transmission could be realized. Results of this work indicate that the purely autoregressive models work well for some types of image textures, but that for the textures studied the ARIMA model is not particularly suitable. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Mar 01, 1987
- Accession Number
- ADA183144
Entities
People
- Steven C. Rathmanner
Organizations
- Naval Postgraduate School