A Global Approach to Image Texture Analysis

• Main Author: Holyer, R. J.
• Other Authors: NAVAL OCEAN RESEARCH AND DEVELOPMENT ACTIVITY STENNIS SPACE CENTER MS
• Format: Text
• Language: English
• Published: 1990
• Subjects: Snow; Ice and Permafrost; Cybernetics; *IMAGE PROCESSING; *TEXTURE; *SEA ICE; SPATIAL DISTRIBUTION; GLOBAL; EXPERIMENTAL DATA; AIRBORNE; PASSIVE SYSTEMS; ACCURACY; MIXTURES; VARIATIONS; ICE; SCALE; IMAGES; REMOTE DETECTORS; MICROWAVE FREQUENCY; ARCTIC REGIONS; ARCTIC OCEAN; MOSAICS(IMAGES); REMOTE SENSING; TEXTURE ANALYSIS; PE62435N; Arctic; Arctic Ocean; Ice; permafrost; Sea ice
• Online Access: http://www.dtic.mil/docs/citations/ADA223532; http://oai.dtic.mil/oai/oai?&verb=getRecord&metadataPrefix=html&identifier=ADA223532

This new approach is based on linear unmixing of texture measures calculated over an entire image (called a global approach), as opposed to most present texture analysis techniques that compute texture over small neighborhoods (called a local approach). The new global paradigm is appropriate for images where spatial scales of the texture variability are large with respect to the pixel spacing, thereby making the local approach ineffective. Airborne passive microwave imagery of Arctic sea ice contain textures that vary with ice-type. These ice textures are of the type best treated by the global approach. Sea-ice imagery is used as test data to evaluate the global techniques that are developed. Pure, single ice-type images; synthetic mixtures formed by mosaicking pure ice-type subimages in known proportions; and naturally occurring mixture images are analyzed. Proportions of first-year, second-year, and multiyear ice within mixture images are retrieved with root-mean-square accuracies as low as 0.04 by the new method. While this accuracy is adequate to be useful in many Arctic studies, the global technique seems promising for many other remote sensing and general image processing applications. Image processing.